Instructions to use mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("feature-extraction", model="mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# Load model directly from transformers import AutoTokenizer, AutoModel tokenizer = AutoTokenizer.from_pretrained("mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless") model = AutoModel.from_pretrained("mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless", device_map="auto") messages = [ {"role": "user", "content": "Who are you?"}, ] inputs = tokenizer.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=40) print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless with llama.cpp:
Install (macOS, Linux)
curl -LsSf https://llama.app/install.sh | sh # Start a local OpenAI-compatible server with a web UI: llama serve -hf mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless:Q8_0 # Run inference directly in the terminal: llama cli -hf mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless:Q8_0
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless:Q8_0 # Run inference directly in the terminal: llama cli -hf mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless:Q8_0
Use pre-built binary
# Download pre-built binary from: # https://github.com/ggerganov/llama.cpp/releases # Start a local OpenAI-compatible server with a web UI: ./llama-server -hf mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless:Q8_0 # Run inference directly in the terminal: ./llama-cli -hf mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless:Q8_0
Build from source code
git clone https://github.com/ggerganov/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-server llama-cli # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama-server -hf mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless:Q8_0 # Run inference directly in the terminal: ./build/bin/llama-cli -hf mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless:Q8_0
Use Docker
docker model run hf.co/mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless:Q8_0
- LM Studio
- Jan
- Ollama
How to use mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless with Ollama:
ollama run hf.co/mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless:Q8_0
- Unsloth Desktop
- Pi
How to use mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless with Pi:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless:Q8_0
Configure the model in Pi
# Install Pi: npm install -g @earendil-works/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "llama-cpp": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless:Q8_0" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Docker Model Runner
How to use mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless with Docker Model Runner:
docker model run hf.co/mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless:Q8_0
- Lemonade
How to use mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless:Q8_0
Run and chat with the model
lemonade run user.qwen3.5-4l-vocab40k-en-ko-headless-Q8_0
List all available models
lemonade list
- Hermes Agent
How to use mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless with Hermes Agent:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless:Q8_0
Configure Hermes
# Install Hermes: curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash hermes setup # Point Hermes at the local server: hermes config set model.provider custom hermes config set model.base_url http://127.0.0.1:8080/v1 hermes config set model.default mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless:Q8_0
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless with OpenClaw:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless:Q8_0
Configure OpenClaw
# Install OpenClaw: npm install -g openclaw@latest # Register the local server and set it as the default model: openclaw onboard --non-interactive --mode local \ --auth-choice custom-api-key \ --custom-base-url http://127.0.0.1:8080/v1 \ --custom-model-id "mp-juuuns/qwen3.5-4l-vocab40k-en-ko-headless:Q8_0" \ --custom-provider-id llama-cpp \ --custom-compatibility openai \ --custom-text-input \ --accept-risk \ --skip-health
Run OpenClaw
openclaw agent --local --agent main --message "Hello from Hugging Face"
Put the full metric comparison at the top; strip local paths; drop stale manifests
Browse filesComparison: the card never placed this root against the field. It now does, at the top, on every metric the protocol measured. Quality rank 3 of 65 (a six-arm tie inside one median seed SD), latency 52/65, energy 52/65, RSS 40/65, file 31/65. New figure 10 shows all 65 arms per metric.
Local paths: seven manifests carried /home/mp_ubuntu/Projects/... absolute paths into a public repository. 24 occurrences rewritten to repo-relative; no other content changed.
Deleted, all stale or wrong:
release_manifest.json - asserts the OLD repo id and the OLD root's weight sha256
benchmark/summary.json - superseded-lineage numbers; also claims a test split this repo
benchmark/semeval-transfer-summary.json does not have. Byte-identical triplicate.
benchmark/figures/source_summary.json
benchmark/figures/manifest.json - schema standalone4l-v128k-figures-v1, describes a
benchmark/figures/cross_task/manifest.json figure set the current card does not use
benchmark/figures/cross_task/source_summary.json
- .gitattributes +1 -0
- benchmark/figures/10_root_on_every_metric.png +3 -0
- benchmark/figures/cross_task/manifest.json +0 -158
- benchmark/figures/cross_task/source_summary.json +0 -972
- benchmark/figures/fig10.py +92 -0
- benchmark/figures/manifest.json +0 -14
- benchmark/figures/source_summary.json +0 -891
- benchmark/semeval-transfer-summary.json +0 -891
- benchmark/summary.json +0 -891
- finalization_manifest.json +2 -2
- provenance/base/finalization_manifest.json +2 -2
- provenance/base/training_manifest.json +4 -4
- provenance/task_agnostic_stages/24to8.json +4 -4
- provenance/task_agnostic_stages/6to4.json +4 -4
- provenance/task_agnostic_stages/8to6.json +4 -4
- release_manifest.json +0 -675
- training_manifest.json +4 -4
|
@@ -11,3 +11,4 @@ benchmark/figures/07_runtime_specialization.png filter=lfs diff=lfs merge=lfs -t
|
|
| 11 |
benchmark/figures/08_long_context_window.png filter=lfs diff=lfs merge=lfs -text
|
| 12 |
docker/bundle/model/classifier-q8_0.gguf filter=lfs diff=lfs merge=lfs -text
|
| 13 |
docker/bundle/runtime_source.tar.gz filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 11 |
benchmark/figures/08_long_context_window.png filter=lfs diff=lfs merge=lfs -text
|
| 12 |
docker/bundle/model/classifier-q8_0.gguf filter=lfs diff=lfs merge=lfs -text
|
| 13 |
docker/bundle/runtime_source.tar.gz filter=lfs diff=lfs merge=lfs -text
|
| 14 |
+
benchmark/figures/10_root_on_every_metric.png filter=lfs diff=lfs merge=lfs -text
|
|
Git LFS Details
|
|
@@ -1,158 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"figures": [
|
| 3 |
-
"05_why_4l_edge_tradeoff",
|
| 4 |
-
"06_malint_4l_quality",
|
| 5 |
-
"07_malint_quality_resource"
|
| 6 |
-
],
|
| 7 |
-
"files_before_manifest": [
|
| 8 |
-
{
|
| 9 |
-
"bytes": 5442,
|
| 10 |
-
"path": "05_why_4l_edge_tradeoff.csv",
|
| 11 |
-
"sha256": "afd203646513c49140a25ef2a562a064286a4fa5efb495e6a73f83f0106e7933"
|
| 12 |
-
},
|
| 13 |
-
{
|
| 14 |
-
"bytes": 27224,
|
| 15 |
-
"path": "05_why_4l_edge_tradeoff.pdf",
|
| 16 |
-
"sha256": "22f531ea586f9c79ac8f114345cec56e74ea5ed97aa3726f7310018666ae6312"
|
| 17 |
-
},
|
| 18 |
-
{
|
| 19 |
-
"bytes": 178965,
|
| 20 |
-
"path": "05_why_4l_edge_tradeoff.png",
|
| 21 |
-
"sha256": "cdc0a8a98cd13ad35a68b980d2e0750e2424b8932f7d27dc0ecd548fed823c74"
|
| 22 |
-
},
|
| 23 |
-
{
|
| 24 |
-
"bytes": 2118,
|
| 25 |
-
"path": "06_malint_4l_quality.csv",
|
| 26 |
-
"sha256": "25848237b9c77596a8c4b6224fa111f86c816e065804ccfdd6a4fdf6c63ef8d1"
|
| 27 |
-
},
|
| 28 |
-
{
|
| 29 |
-
"bytes": 26351,
|
| 30 |
-
"path": "06_malint_4l_quality.pdf",
|
| 31 |
-
"sha256": "9a1ae78b4e5871f5d5c5321db6c5d1fe0bf349a7869977b2649b44ccea633d04"
|
| 32 |
-
},
|
| 33 |
-
{
|
| 34 |
-
"bytes": 181392,
|
| 35 |
-
"path": "06_malint_4l_quality.png",
|
| 36 |
-
"sha256": "2c5b22415d311ca5fa2b198d55b77f3570b57595d01f0b5031abde59ffc3e9e6"
|
| 37 |
-
},
|
| 38 |
-
{
|
| 39 |
-
"bytes": 1800,
|
| 40 |
-
"path": "07_malint_quality_resource.csv",
|
| 41 |
-
"sha256": "aac7d3400cf8dcb3ac8b6fc317c27ded5343231c93476997011a3bb56b5e4d6c"
|
| 42 |
-
},
|
| 43 |
-
{
|
| 44 |
-
"bytes": 29309,
|
| 45 |
-
"path": "07_malint_quality_resource.pdf",
|
| 46 |
-
"sha256": "498338991976e64038096a12361f63830ced3b9c6eea3e1bdaed8d0e4efac606"
|
| 47 |
-
},
|
| 48 |
-
{
|
| 49 |
-
"bytes": 163562,
|
| 50 |
-
"path": "07_malint_quality_resource.png",
|
| 51 |
-
"sha256": "1cb13cc695f3d59ae9ce75599b86867da624d6e0aae0897e0db8686f6db984ba"
|
| 52 |
-
},
|
| 53 |
-
{
|
| 54 |
-
"bytes": 868,
|
| 55 |
-
"path": "README.md",
|
| 56 |
-
"sha256": "f3e79965f14ec0fd20d11c361d25c3a3abc4c9a20b5dbce9fac0660bb12789f0"
|
| 57 |
-
},
|
| 58 |
-
{
|
| 59 |
-
"bytes": 29752,
|
| 60 |
-
"path": "source_summary.json",
|
| 61 |
-
"sha256": "20ea292f85757440d8641199003c47674deb0c6aacea0ef8f9f5fcd2480bb5d0"
|
| 62 |
-
}
|
| 63 |
-
],
|
| 64 |
-
"inputs": [
|
| 65 |
-
{
|
| 66 |
-
"bytes": 58857,
|
| 67 |
-
"path": "results/malint_qwen35_4l_repeated_cv/20260814-r3x3-terminal-v1/aggregate-v1/aggregate_result.json",
|
| 68 |
-
"role": "malint_repeated_aggregate",
|
| 69 |
-
"sha256": "e3d659c2f5aca645f040d03c57e4effa9eb2616f622aceb503c71236b6b01efa"
|
| 70 |
-
},
|
| 71 |
-
{
|
| 72 |
-
"bytes": 325527,
|
| 73 |
-
"path": "results/malint_qwen35_4l_repeated_cv/20260814-r3x3-terminal-v1/aggregate-v1/aggregate_oof_predictions.npz",
|
| 74 |
-
"role": "malint_repeated_oof",
|
| 75 |
-
"sha256": "a1d7827ebe4beecb49676ca42e0c986c5f4101c394d55c9965161481c347a3f1"
|
| 76 |
-
},
|
| 77 |
-
{
|
| 78 |
-
"bytes": 15407,
|
| 79 |
-
"path": "results/malint_qwen35_4l_resource_benchmark/20260814-matched-bf16-v2/resource_benchmark.json",
|
| 80 |
-
"role": "malint_resource_aggregate",
|
| 81 |
-
"sha256": "5337e44c907a05590c0ea3da6a0760baff22c3f930ed81c069c257060fd2e725"
|
| 82 |
-
},
|
| 83 |
-
{
|
| 84 |
-
"bytes": 3144,
|
| 85 |
-
"path": "results/final_broad_benchmark/20260729-v1/report/quality_matrix.csv",
|
| 86 |
-
"role": "propaganda_quality_matrix",
|
| 87 |
-
"sha256": "c6e316f7081e2a6dcda1ec4e95b57cdcc357ae49ed8b3f5a876c5b51e0df722f"
|
| 88 |
-
},
|
| 89 |
-
{
|
| 90 |
-
"bytes": 115661,
|
| 91 |
-
"path": "results/final_broad_benchmark/20260729-v1/trainable/qwen35-full24l-commonhead/seed41/report.json",
|
| 92 |
-
"role": "propaganda_seed_report_01",
|
| 93 |
-
"sha256": "1c572fb728755dbde0460e6b8eb9a84d078b128f8a42a9148a17c40fc0b27d82"
|
| 94 |
-
},
|
| 95 |
-
{
|
| 96 |
-
"bytes": 116024,
|
| 97 |
-
"path": "results/final_broad_benchmark/20260729-v1/trainable/qwen35-full24l-commonhead/seed42/report.json",
|
| 98 |
-
"role": "propaganda_seed_report_02",
|
| 99 |
-
"sha256": "055edfee92e39928aa9577f239c5b0389f3dda375a96fc69a2c000a0ed6edc36"
|
| 100 |
-
},
|
| 101 |
-
{
|
| 102 |
-
"bytes": 115639,
|
| 103 |
-
"path": "results/final_broad_benchmark/20260729-v1/trainable/qwen35-full24l-commonhead/seed43-retry2/report.json",
|
| 104 |
-
"role": "propaganda_seed_report_03",
|
| 105 |
-
"sha256": "84b2c316556db4aa5fdf7c08433a4808a5fe9ef6f620da6472f4f2f73dee2e4b"
|
| 106 |
-
},
|
| 107 |
-
{
|
| 108 |
-
"bytes": 114924,
|
| 109 |
-
"path": "results/final_broad_benchmark/20260729-v1/trainable/qwen35-c8q8l-commonhead/seed41/report.json",
|
| 110 |
-
"role": "propaganda_seed_report_04",
|
| 111 |
-
"sha256": "9234258589ffb24eaa92859639a66aeb3e55190f8aa82f798d88e791a282bf9c"
|
| 112 |
-
},
|
| 113 |
-
{
|
| 114 |
-
"bytes": 115800,
|
| 115 |
-
"path": "results/final_broad_benchmark/20260729-v1/trainable/qwen35-c8q8l-commonhead/seed42/report.json",
|
| 116 |
-
"role": "propaganda_seed_report_05",
|
| 117 |
-
"sha256": "8161dd1378081de9960136fddab800994529aa3c754674ca41bb4aa71515b3c6"
|
| 118 |
-
},
|
| 119 |
-
{
|
| 120 |
-
"bytes": 116039,
|
| 121 |
-
"path": "results/final_broad_benchmark/20260729-v1/trainable/qwen35-c8q8l-commonhead/seed43/report.json",
|
| 122 |
-
"role": "propaganda_seed_report_06",
|
| 123 |
-
"sha256": "ce0c76a2aa97a5b7c8cdf820af0bd4f8a26bf50d460612ca1db30bd497ca3d9d"
|
| 124 |
-
},
|
| 125 |
-
{
|
| 126 |
-
"bytes": 115985,
|
| 127 |
-
"path": "results/final_broad_benchmark/20260729-v1/trainable/qwen35-standalone4l-commonhead/seed41/report.json",
|
| 128 |
-
"role": "propaganda_seed_report_07",
|
| 129 |
-
"sha256": "ec5314140fff0bdd46e87ee00ad67bee7e61d82c7cfa34add1510dd7d5bdb3c0"
|
| 130 |
-
},
|
| 131 |
-
{
|
| 132 |
-
"bytes": 115972,
|
| 133 |
-
"path": "results/final_broad_benchmark/20260729-v1/trainable/qwen35-standalone4l-commonhead/seed42/report.json",
|
| 134 |
-
"role": "propaganda_seed_report_08",
|
| 135 |
-
"sha256": "e5c45a7dc41bebf4550da7c5763d0759c6d0922da9efda4048564c11c6d2f3af"
|
| 136 |
-
},
|
| 137 |
-
{
|
| 138 |
-
"bytes": 116177,
|
| 139 |
-
"path": "results/final_broad_benchmark/20260729-v1/trainable/qwen35-standalone4l-commonhead/seed43/report.json",
|
| 140 |
-
"role": "propaganda_seed_report_09",
|
| 141 |
-
"sha256": "e05bc2146d720706780fc45a1fb05911d5c2926caa0bcb12ffead946028eef0d"
|
| 142 |
-
},
|
| 143 |
-
{
|
| 144 |
-
"bytes": 827,
|
| 145 |
-
"path": "results/standalone4l_base/20260731-v1/huggingface_release_v3/benchmark/compression_ladder.csv",
|
| 146 |
-
"role": "task_free_compression_ladder",
|
| 147 |
-
"sha256": "2c544c3907c4f592d724f658850b8c6b9634845cb5b50a66ebd2ea6c5d48ce0b"
|
| 148 |
-
},
|
| 149 |
-
{
|
| 150 |
-
"bytes": 13963,
|
| 151 |
-
"path": "results/unoq_lineage_matrix/20260731-v1/analysis/analysis.json",
|
| 152 |
-
"role": "uno_q_depth_analysis",
|
| 153 |
-
"sha256": "c899a19e4433e3a40547a11cfcfc3b0a7b6f00c436aa8c1d742a5dc0399bf649"
|
| 154 |
-
}
|
| 155 |
-
],
|
| 156 |
-
"schema": "qwen35-4l-cross-task-base-card-figures-manifest-v1",
|
| 157 |
-
"status": "complete"
|
| 158 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@@ -1,972 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"claim_boundaries": [
|
| 3 |
-
"The controlled UNO Q 4L point is a historical early-path proxy, not the final common-head quality endpoint.",
|
| 4 |
-
"The propaganda 24L/8L/4L endpoints differ in lineage and vocabulary; their comparison is descriptive, not a pure depth ablation.",
|
| 5 |
-
"All Qwen MALINT student arms are 4L; MALINT supports a 4L interface result, not 4L-versus-8L/24L depth superiority.",
|
| 6 |
-
"The MALINT result is post-selection internal repeated-fold evidence, not untouched held-out confirmation or SOTA evidence.",
|
| 7 |
-
"Mean+KD incurred measured CUDA-forward latency and peak-allocation increases; it is not claimed to be faster.",
|
| 8 |
-
"No ambient GPU utilization value is used as model occupancy."
|
| 9 |
-
],
|
| 10 |
-
"malint_quality": [
|
| 11 |
-
{
|
| 12 |
-
"arm": "rightmost_hard",
|
| 13 |
-
"arm_label": "Rightmost + hard",
|
| 14 |
-
"layers": 4,
|
| 15 |
-
"mean": 0.522182806412052,
|
| 16 |
-
"metric": "weighted_f1",
|
| 17 |
-
"n_repeats": 3,
|
| 18 |
-
"role": "qwen_4l_student",
|
| 19 |
-
"sample_sd": 0.014974460565445307,
|
| 20 |
-
"threshold": 0.3
|
| 21 |
-
},
|
| 22 |
-
{
|
| 23 |
-
"arm": "rightmost_hard",
|
| 24 |
-
"arm_label": "Rightmost + hard",
|
| 25 |
-
"layers": 4,
|
| 26 |
-
"mean": 0.5109624469249431,
|
| 27 |
-
"metric": "macro_f1",
|
| 28 |
-
"n_repeats": 3,
|
| 29 |
-
"role": "qwen_4l_student",
|
| 30 |
-
"sample_sd": 0.016275524962473233,
|
| 31 |
-
"threshold": 0.3
|
| 32 |
-
},
|
| 33 |
-
{
|
| 34 |
-
"arm": "rightmost_hard",
|
| 35 |
-
"arm_label": "Rightmost + hard",
|
| 36 |
-
"layers": 4,
|
| 37 |
-
"mean": 0.5320256865023298,
|
| 38 |
-
"metric": "micro_f1",
|
| 39 |
-
"n_repeats": 3,
|
| 40 |
-
"role": "qwen_4l_student",
|
| 41 |
-
"sample_sd": 0.010757300158510568,
|
| 42 |
-
"threshold": 0.3
|
| 43 |
-
},
|
| 44 |
-
{
|
| 45 |
-
"arm": "rightmost_hard",
|
| 46 |
-
"arm_label": "Rightmost + hard",
|
| 47 |
-
"layers": 4,
|
| 48 |
-
"mean": 0.6403030303030303,
|
| 49 |
-
"metric": "subset_exact_match",
|
| 50 |
-
"n_repeats": 3,
|
| 51 |
-
"role": "qwen_4l_student",
|
| 52 |
-
"sample_sd": 0.003674047167570365,
|
| 53 |
-
"threshold": 0.3
|
| 54 |
-
},
|
| 55 |
-
{
|
| 56 |
-
"arm": "rightmost_hard",
|
| 57 |
-
"arm_label": "Rightmost + hard",
|
| 58 |
-
"layers": 4,
|
| 59 |
-
"mean": 0.8746666666666667,
|
| 60 |
-
"metric": "label_accuracy",
|
| 61 |
-
"n_repeats": 3,
|
| 62 |
-
"role": "qwen_4l_student",
|
| 63 |
-
"sample_sd": 0.0012106051124507965,
|
| 64 |
-
"threshold": 0.3
|
| 65 |
-
},
|
| 66 |
-
{
|
| 67 |
-
"arm": "mean_hard",
|
| 68 |
-
"arm_label": "Mean pool + hard",
|
| 69 |
-
"layers": 4,
|
| 70 |
-
"mean": 0.6118049035859879,
|
| 71 |
-
"metric": "weighted_f1",
|
| 72 |
-
"n_repeats": 3,
|
| 73 |
-
"role": "qwen_4l_student",
|
| 74 |
-
"sample_sd": 0.008860176348616246,
|
| 75 |
-
"threshold": 0.3
|
| 76 |
-
},
|
| 77 |
-
{
|
| 78 |
-
"arm": "mean_hard",
|
| 79 |
-
"arm_label": "Mean pool + hard",
|
| 80 |
-
"layers": 4,
|
| 81 |
-
"mean": 0.6055064323204021,
|
| 82 |
-
"metric": "macro_f1",
|
| 83 |
-
"n_repeats": 3,
|
| 84 |
-
"role": "qwen_4l_student",
|
| 85 |
-
"sample_sd": 0.007533985719586246,
|
| 86 |
-
"threshold": 0.3
|
| 87 |
-
},
|
| 88 |
-
{
|
| 89 |
-
"arm": "mean_hard",
|
| 90 |
-
"arm_label": "Mean pool + hard",
|
| 91 |
-
"layers": 4,
|
| 92 |
-
"mean": 0.6176900341883843,
|
| 93 |
-
"metric": "micro_f1",
|
| 94 |
-
"n_repeats": 3,
|
| 95 |
-
"role": "qwen_4l_student",
|
| 96 |
-
"sample_sd": 0.007461802499370809,
|
| 97 |
-
"threshold": 0.3
|
| 98 |
-
},
|
| 99 |
-
{
|
| 100 |
-
"arm": "mean_hard",
|
| 101 |
-
"arm_label": "Mean pool + hard",
|
| 102 |
-
"layers": 4,
|
| 103 |
-
"mean": 0.6696969696969697,
|
| 104 |
-
"metric": "subset_exact_match",
|
| 105 |
-
"n_repeats": 3,
|
| 106 |
-
"role": "qwen_4l_student",
|
| 107 |
-
"sample_sd": 0.005169309730070291,
|
| 108 |
-
"threshold": 0.3
|
| 109 |
-
},
|
| 110 |
-
{
|
| 111 |
-
"arm": "mean_hard",
|
| 112 |
-
"arm_label": "Mean pool + hard",
|
| 113 |
-
"layers": 4,
|
| 114 |
-
"mean": 0.8876363636363637,
|
| 115 |
-
"metric": "label_accuracy",
|
| 116 |
-
"n_repeats": 3,
|
| 117 |
-
"role": "qwen_4l_student",
|
| 118 |
-
"sample_sd": 0.0017533910474532954,
|
| 119 |
-
"threshold": 0.3
|
| 120 |
-
},
|
| 121 |
-
{
|
| 122 |
-
"arm": "mean_kd",
|
| 123 |
-
"arm_label": "Mean pool + KD",
|
| 124 |
-
"layers": 4,
|
| 125 |
-
"mean": 0.6346533875063399,
|
| 126 |
-
"metric": "weighted_f1",
|
| 127 |
-
"n_repeats": 3,
|
| 128 |
-
"role": "qwen_4l_student",
|
| 129 |
-
"sample_sd": 0.011966996735426847,
|
| 130 |
-
"threshold": 0.3
|
| 131 |
-
},
|
| 132 |
-
{
|
| 133 |
-
"arm": "mean_kd",
|
| 134 |
-
"arm_label": "Mean pool + KD",
|
| 135 |
-
"layers": 4,
|
| 136 |
-
"mean": 0.6281768437686408,
|
| 137 |
-
"metric": "macro_f1",
|
| 138 |
-
"n_repeats": 3,
|
| 139 |
-
"role": "qwen_4l_student",
|
| 140 |
-
"sample_sd": 0.014225316120600004,
|
| 141 |
-
"threshold": 0.3
|
| 142 |
-
},
|
| 143 |
-
{
|
| 144 |
-
"arm": "mean_kd",
|
| 145 |
-
"arm_label": "Mean pool + KD",
|
| 146 |
-
"layers": 4,
|
| 147 |
-
"mean": 0.6389632888388834,
|
| 148 |
-
"metric": "micro_f1",
|
| 149 |
-
"n_repeats": 3,
|
| 150 |
-
"role": "qwen_4l_student",
|
| 151 |
-
"sample_sd": 0.009850482866928375,
|
| 152 |
-
"threshold": 0.3
|
| 153 |
-
},
|
| 154 |
-
{
|
| 155 |
-
"arm": "mean_kd",
|
| 156 |
-
"arm_label": "Mean pool + KD",
|
| 157 |
-
"layers": 4,
|
| 158 |
-
"mean": 0.6557575757575758,
|
| 159 |
-
"metric": "subset_exact_match",
|
| 160 |
-
"n_repeats": 3,
|
| 161 |
-
"role": "qwen_4l_student",
|
| 162 |
-
"sample_sd": 0.003192622349349309,
|
| 163 |
-
"threshold": 0.3
|
| 164 |
-
},
|
| 165 |
-
{
|
| 166 |
-
"arm": "mean_kd",
|
| 167 |
-
"arm_label": "Mean pool + KD",
|
| 168 |
-
"layers": 4,
|
| 169 |
-
"mean": 0.8757575757575757,
|
| 170 |
-
"metric": "label_accuracy",
|
| 171 |
-
"n_repeats": 3,
|
| 172 |
-
"role": "qwen_4l_student",
|
| 173 |
-
"sample_sd": 0.0020436018545560413,
|
| 174 |
-
"threshold": 0.3
|
| 175 |
-
},
|
| 176 |
-
{
|
| 177 |
-
"arm": "teacher",
|
| 178 |
-
"arm_label": "RoBERTa teacher",
|
| 179 |
-
"layers": "",
|
| 180 |
-
"mean": 0.6046800630077942,
|
| 181 |
-
"metric": "weighted_f1",
|
| 182 |
-
"n_repeats": 3,
|
| 183 |
-
"role": "external_teacher",
|
| 184 |
-
"sample_sd": 0.01579900095351724,
|
| 185 |
-
"threshold": 0.3
|
| 186 |
-
},
|
| 187 |
-
{
|
| 188 |
-
"arm": "teacher",
|
| 189 |
-
"arm_label": "RoBERTa teacher",
|
| 190 |
-
"layers": "",
|
| 191 |
-
"mean": 0.5955721586042237,
|
| 192 |
-
"metric": "macro_f1",
|
| 193 |
-
"n_repeats": 3,
|
| 194 |
-
"role": "external_teacher",
|
| 195 |
-
"sample_sd": 0.01637303032583919,
|
| 196 |
-
"threshold": 0.3
|
| 197 |
-
},
|
| 198 |
-
{
|
| 199 |
-
"arm": "teacher",
|
| 200 |
-
"arm_label": "RoBERTa teacher",
|
| 201 |
-
"layers": "",
|
| 202 |
-
"mean": 0.6046188233688233,
|
| 203 |
-
"metric": "micro_f1",
|
| 204 |
-
"n_repeats": 3,
|
| 205 |
-
"role": "external_teacher",
|
| 206 |
-
"sample_sd": 0.01776761734134435,
|
| 207 |
-
"threshold": 0.3
|
| 208 |
-
},
|
| 209 |
-
{
|
| 210 |
-
"arm": "teacher",
|
| 211 |
-
"arm_label": "RoBERTa teacher",
|
| 212 |
-
"layers": "",
|
| 213 |
-
"mean": 0.62,
|
| 214 |
-
"metric": "subset_exact_match",
|
| 215 |
-
"n_repeats": 3,
|
| 216 |
-
"role": "external_teacher",
|
| 217 |
-
"sample_sd": 0.005961307749365459,
|
| 218 |
-
"threshold": 0.3
|
| 219 |
-
},
|
| 220 |
-
{
|
| 221 |
-
"arm": "teacher",
|
| 222 |
-
"arm_label": "RoBERTa teacher",
|
| 223 |
-
"layers": "",
|
| 224 |
-
"mean": 0.8507878787878788,
|
| 225 |
-
"metric": "label_accuracy",
|
| 226 |
-
"n_repeats": 3,
|
| 227 |
-
"role": "external_teacher",
|
| 228 |
-
"sample_sd": 0.01087469324477906,
|
| 229 |
-
"threshold": 0.3
|
| 230 |
-
}
|
| 231 |
-
],
|
| 232 |
-
"malint_quality_contract": {
|
| 233 |
-
"claim_boundary": "Post-selection internal repeated-fold robustness and matched ablation only. This result is not untouched confirmation, official held-out evidence, all-label superiority, or SOTA evidence.",
|
| 234 |
-
"contrasts": {
|
| 235 |
-
"conditional_kd_mean_kd_minus_mean_hard": {
|
| 236 |
-
"corrected_resampled_t_interval": {
|
| 237 |
-
"confidence": 0.95,
|
| 238 |
-
"critical_t": 2.306004135204166,
|
| 239 |
-
"degrees_of_freedom": 8,
|
| 240 |
-
"formula": "sqrt((1/n + mean(n_eval/n_train)) * sample_variance)",
|
| 241 |
-
"interval": [
|
| 242 |
-
-0.0015478418507784523,
|
| 243 |
-
0.05059146195651548
|
| 244 |
-
],
|
| 245 |
-
"mean": 0.024521810052868513,
|
| 246 |
-
"mean_eval_train_ratio": 0.5000006195533516,
|
| 247 |
-
"n": 9,
|
| 248 |
-
"sample_variance": 0.00020913642539219796,
|
| 249 |
-
"standard_error": 0.011305119321192738,
|
| 250 |
-
"variance_correction": 0.6111117306644627
|
| 251 |
-
},
|
| 252 |
-
"fold_deltas": [
|
| 253 |
-
0.04524702943141845,
|
| 254 |
-
0.017956307877355804,
|
| 255 |
-
-0.0053068243793217285,
|
| 256 |
-
0.015864642448092825,
|
| 257 |
-
0.022064206421293142,
|
| 258 |
-
0.03213041489613799,
|
| 259 |
-
0.02962871459506977,
|
| 260 |
-
0.026719118927120977,
|
| 261 |
-
0.03639268025864939
|
| 262 |
-
],
|
| 263 |
-
"fold_mean": 0.024521810052868513,
|
| 264 |
-
"fold_sample_standard_deviation": 0.014461549895920491,
|
| 265 |
-
"name": "conditional_kd_mean_kd_minus_mean_hard",
|
| 266 |
-
"per_label_mean_delta": {
|
| 267 |
-
"CPV": 0.0006438268748058898,
|
| 268 |
-
"PASV": 0.027106289374744647,
|
| 269 |
-
"PSSA": 0.058741661936162225,
|
| 270 |
-
"UCPI": 0.0247139382339762,
|
| 271 |
-
"UIOA": 0.0021463408215049995
|
| 272 |
-
},
|
| 273 |
-
"per_label_repeat_deltas": [
|
| 274 |
-
[
|
| 275 |
-
-0.01842723004694835,
|
| 276 |
-
0.04865531704909942,
|
| 277 |
-
0.015803620583568212,
|
| 278 |
-
-0.013921424268790505,
|
| 279 |
-
0.03679497238444562
|
| 280 |
-
],
|
| 281 |
-
[
|
| 282 |
-
0.025183733463988223,
|
| 283 |
-
0.070038654494613,
|
| 284 |
-
-0.0026296349009481412,
|
| 285 |
-
0.021790722761596548,
|
| 286 |
-
0.0004450534649754756
|
| 287 |
-
],
|
| 288 |
-
[
|
| 289 |
-
-0.004825022792622202,
|
| 290 |
-
0.05753101426477425,
|
| 291 |
-
-0.006734963218105072,
|
| 292 |
-
0.0734495696314279,
|
| 293 |
-
0.0369017888525075
|
| 294 |
-
]
|
| 295 |
-
],
|
| 296 |
-
"positive_fold_tail": {
|
| 297 |
-
"null_probability": 0.5,
|
| 298 |
-
"one_sided_probability_at_least_observed": 0.01953125,
|
| 299 |
-
"positive": 8,
|
| 300 |
-
"total": 9
|
| 301 |
-
},
|
| 302 |
-
"repeat_oof_deltas": [
|
| 303 |
-
0.018545164631776245,
|
| 304 |
-
0.020840853227636957,
|
| 305 |
-
0.029159433901642884
|
| 306 |
-
],
|
| 307 |
-
"repeat_oof_t_interval": {
|
| 308 |
-
"confidence": 0.95,
|
| 309 |
-
"critical_t": 4.302652729749462,
|
| 310 |
-
"degrees_of_freedom": 2,
|
| 311 |
-
"interval": [
|
| 312 |
-
0.00897537793561895,
|
| 313 |
-
0.03672158990508511
|
| 314 |
-
],
|
| 315 |
-
"mean": 0.02284848392035203,
|
| 316 |
-
"n": 3,
|
| 317 |
-
"sample_standard_deviation": 0.005584676694496962,
|
| 318 |
-
"standard_error": 0.0032243145929048505
|
| 319 |
-
}
|
| 320 |
-
},
|
| 321 |
-
"mean_kd_minus_teacher": {
|
| 322 |
-
"corrected_resampled_t_interval": {
|
| 323 |
-
"confidence": 0.95,
|
| 324 |
-
"critical_t": 2.306004135204166,
|
| 325 |
-
"degrees_of_freedom": 8,
|
| 326 |
-
"formula": "sqrt((1/n + mean(n_eval/n_train)) * sample_variance)",
|
| 327 |
-
"interval": [
|
| 328 |
-
-0.03180974553628697,
|
| 329 |
-
0.08853586985785872
|
| 330 |
-
],
|
| 331 |
-
"mean": 0.02836306216078588,
|
| 332 |
-
"mean_eval_train_ratio": 0.5000006195533516,
|
| 333 |
-
"n": 9,
|
| 334 |
-
"sample_variance": 0.0011141913151793519,
|
| 335 |
-
"standard_error": 0.026093972156622066,
|
| 336 |
-
"variance_correction": 0.6111117306644627
|
| 337 |
-
},
|
| 338 |
-
"fold_deltas": [
|
| 339 |
-
-0.03162698747672954,
|
| 340 |
-
0.06379890775996899,
|
| 341 |
-
0.056964918230746675,
|
| 342 |
-
0.04693837159160652,
|
| 343 |
-
0.001345287770746495,
|
| 344 |
-
0.014058020214096523,
|
| 345 |
-
0.00030085368163068615,
|
| 346 |
-
0.05808798200431842,
|
| 347 |
-
0.04540020567068814
|
| 348 |
-
],
|
| 349 |
-
"fold_mean": 0.02836306216078588,
|
| 350 |
-
"fold_sample_standard_deviation": 0.0333795044178213,
|
| 351 |
-
"name": "mean_kd_minus_teacher",
|
| 352 |
-
"per_label_mean_delta": {
|
| 353 |
-
"CPV": 0.03245133981428311,
|
| 354 |
-
"PASV": 0.04215462872506418,
|
| 355 |
-
"PSSA": 0.030550793916265178,
|
| 356 |
-
"UCPI": 0.006157663285133859,
|
| 357 |
-
"UIOA": 0.05170900008133931
|
| 358 |
-
},
|
| 359 |
-
"per_label_repeat_deltas": [
|
| 360 |
-
[
|
| 361 |
-
0.010081541882876188,
|
| 362 |
-
0.09069144184384492,
|
| 363 |
-
0.058604490711147705,
|
| 364 |
-
-0.005178527872112815,
|
| 365 |
-
0.0012970008399624255
|
| 366 |
-
],
|
| 367 |
-
[
|
| 368 |
-
0.05816392692322364,
|
| 369 |
-
-0.013100009776126686,
|
| 370 |
-
0.017979371387083054,
|
| 371 |
-
0.03578643578643581,
|
| 372 |
-
0.01576007056748019
|
| 373 |
-
],
|
| 374 |
-
[
|
| 375 |
-
0.029108550636749497,
|
| 376 |
-
0.014060949681077295,
|
| 377 |
-
0.07854313814578717,
|
| 378 |
-
0.09585597826086956,
|
| 379 |
-
0.0014159184479589637
|
| 380 |
-
]
|
| 381 |
-
],
|
| 382 |
-
"positive_fold_tail": {
|
| 383 |
-
"null_probability": 0.5,
|
| 384 |
-
"one_sided_probability_at_least_observed": 0.01953125,
|
| 385 |
-
"positive": 8,
|
| 386 |
-
"total": 9
|
| 387 |
-
},
|
| 388 |
-
"repeat_oof_deltas": [
|
| 389 |
-
0.032162246558158025,
|
| 390 |
-
0.020416569493259984,
|
| 391 |
-
0.037341157444219175
|
| 392 |
-
],
|
| 393 |
-
"repeat_oof_t_interval": {
|
| 394 |
-
"confidence": 0.95,
|
| 395 |
-
"critical_t": 4.302652729749462,
|
| 396 |
-
"degrees_of_freedom": 2,
|
| 397 |
-
"interval": [
|
| 398 |
-
0.00843082924454409,
|
| 399 |
-
0.05151581975254736
|
| 400 |
-
],
|
| 401 |
-
"mean": 0.029973324498545728,
|
| 402 |
-
"n": 3,
|
| 403 |
-
"sample_standard_deviation": 0.008672021342497449,
|
| 404 |
-
"standard_error": 0.005006793856509082
|
| 405 |
-
}
|
| 406 |
-
},
|
| 407 |
-
"pooling_mean_hard_minus_rightmost_hard": {
|
| 408 |
-
"corrected_resampled_t_interval": {
|
| 409 |
-
"confidence": 0.95,
|
| 410 |
-
"critical_t": 2.306004135204166,
|
| 411 |
-
"degrees_of_freedom": 8,
|
| 412 |
-
"formula": "sqrt((1/n + mean(n_eval/n_train)) * sample_variance)",
|
| 413 |
-
"interval": [
|
| 414 |
-
0.056762831891190316,
|
| 415 |
-
0.12504455147590826
|
| 416 |
-
],
|
| 417 |
-
"mean": 0.0909036916835493,
|
| 418 |
-
"mean_eval_train_ratio": 0.5000006195533516,
|
| 419 |
-
"n": 9,
|
| 420 |
-
"sample_variance": 0.00035868079546522353,
|
| 421 |
-
"standard_error": 0.014805203195932804,
|
| 422 |
-
"variance_correction": 0.6111117306644627
|
| 423 |
-
},
|
| 424 |
-
"fold_deltas": [
|
| 425 |
-
0.06678423059606675,
|
| 426 |
-
0.08178035353463753,
|
| 427 |
-
0.08672762777739151,
|
| 428 |
-
0.09344774812516743,
|
| 429 |
-
0.0878998694863643,
|
| 430 |
-
0.07415570161139795,
|
| 431 |
-
0.11094545718958193,
|
| 432 |
-
0.08708564234043492,
|
| 433 |
-
0.12930659449090132
|
| 434 |
-
],
|
| 435 |
-
"fold_mean": 0.0909036916835493,
|
| 436 |
-
"fold_sample_standard_deviation": 0.01893886996273071,
|
| 437 |
-
"name": "pooling_mean_hard_minus_rightmost_hard",
|
| 438 |
-
"per_label_mean_delta": {
|
| 439 |
-
"CPV": 0.14860808426516173,
|
| 440 |
-
"PASV": 0.09778905085334237,
|
| 441 |
-
"PSSA": 0.07923875253833933,
|
| 442 |
-
"UCPI": 0.055020095369494,
|
| 443 |
-
"UIOA": 0.09206394395095778
|
| 444 |
-
},
|
| 445 |
-
"per_label_repeat_deltas": [
|
| 446 |
-
[
|
| 447 |
-
0.09436936936936935,
|
| 448 |
-
0.07977821265492502,
|
| 449 |
-
0.10417829690385194,
|
| 450 |
-
0.07872402692420255,
|
| 451 |
-
0.03863953001237319
|
| 452 |
-
],
|
| 453 |
-
[
|
| 454 |
-
0.15668016194331985,
|
| 455 |
-
0.04167051934042232,
|
| 456 |
-
0.06934192200557099,
|
| 457 |
-
0.11942791523123497,
|
| 458 |
-
0.06908520415955133
|
| 459 |
-
],
|
| 460 |
-
[
|
| 461 |
-
0.194774721482796,
|
| 462 |
-
0.11626752561967063,
|
| 463 |
-
0.1026716129434504,
|
| 464 |
-
0.09521521040458958,
|
| 465 |
-
0.05733555193655748
|
| 466 |
-
]
|
| 467 |
-
],
|
| 468 |
-
"positive_fold_tail": {
|
| 469 |
-
"null_probability": 0.5,
|
| 470 |
-
"one_sided_probability_at_least_observed": 0.001953125,
|
| 471 |
-
"positive": 9,
|
| 472 |
-
"total": 9
|
| 473 |
-
},
|
| 474 |
-
"repeat_oof_deltas": [
|
| 475 |
-
0.07574451062131737,
|
| 476 |
-
0.08540665760778032,
|
| 477 |
-
0.10771512329271016
|
| 478 |
-
],
|
| 479 |
-
"repeat_oof_t_interval": {
|
| 480 |
-
"confidence": 0.95,
|
| 481 |
-
"critical_t": 4.302652729749462,
|
| 482 |
-
"degrees_of_freedom": 2,
|
| 483 |
-
"interval": [
|
| 484 |
-
0.048890004893968025,
|
| 485 |
-
0.13035418945390387
|
| 486 |
-
],
|
| 487 |
-
"mean": 0.08962209717393595,
|
| 488 |
-
"n": 3,
|
| 489 |
-
"sample_standard_deviation": 0.01639687368671199,
|
| 490 |
-
"standard_error": 0.009466739436891459
|
| 491 |
-
}
|
| 492 |
-
},
|
| 493 |
-
"primary_mean_kd_minus_rightmost_hard": {
|
| 494 |
-
"corrected_resampled_t_interval": {
|
| 495 |
-
"confidence": 0.95,
|
| 496 |
-
"critical_t": 2.306004135204166,
|
| 497 |
-
"degrees_of_freedom": 8,
|
| 498 |
-
"formula": "sqrt((1/n + mean(n_eval/n_train)) * sample_variance)",
|
| 499 |
-
"interval": [
|
| 500 |
-
0.07160480852626178,
|
| 501 |
-
0.15924619494657383
|
| 502 |
-
],
|
| 503 |
-
"mean": 0.1154255017364178,
|
| 504 |
-
"mean_eval_train_ratio": 0.5000006195533516,
|
| 505 |
-
"n": 9,
|
| 506 |
-
"sample_variance": 0.0005909050521203029,
|
| 507 |
-
"standard_error": 0.019002868442938105,
|
| 508 |
-
"variance_correction": 0.6111117306644627
|
| 509 |
-
},
|
| 510 |
-
"fold_deltas": [
|
| 511 |
-
0.1120312600274852,
|
| 512 |
-
0.09973666141199333,
|
| 513 |
-
0.08142080339806979,
|
| 514 |
-
0.10931239057326025,
|
| 515 |
-
0.10996407590765744,
|
| 516 |
-
0.10628611650753594,
|
| 517 |
-
0.1405741717846517,
|
| 518 |
-
0.1138047612675559,
|
| 519 |
-
0.1656992747495507
|
| 520 |
-
],
|
| 521 |
-
"fold_mean": 0.1154255017364178,
|
| 522 |
-
"fold_sample_standard_deviation": 0.024308538666902683,
|
| 523 |
-
"name": "primary_mean_kd_minus_rightmost_hard",
|
| 524 |
-
"per_label_mean_delta": {
|
| 525 |
-
"CPV": 0.14925191113996764,
|
| 526 |
-
"PASV": 0.12489534022808702,
|
| 527 |
-
"PSSA": 0.13798041447450154,
|
| 528 |
-
"UCPI": 0.0797340336034702,
|
| 529 |
-
"UIOA": 0.09421028477246278
|
| 530 |
-
},
|
| 531 |
-
"per_label_repeat_deltas": [
|
| 532 |
-
[
|
| 533 |
-
0.075942139322421,
|
| 534 |
-
0.12843352970402444,
|
| 535 |
-
0.11998191748742015,
|
| 536 |
-
0.06480260265541204,
|
| 537 |
-
0.07543450239681881
|
| 538 |
-
],
|
| 539 |
-
[
|
| 540 |
-
0.18186389540730807,
|
| 541 |
-
0.11170917383503531,
|
| 542 |
-
0.06671228710462285,
|
| 543 |
-
0.14121863799283152,
|
| 544 |
-
0.0695302576245268
|
| 545 |
-
],
|
| 546 |
-
[
|
| 547 |
-
0.1899496986901738,
|
| 548 |
-
0.17379853988444488,
|
| 549 |
-
0.09593664972534532,
|
| 550 |
-
0.16866478003601748,
|
| 551 |
-
0.09423734078906498
|
| 552 |
-
]
|
| 553 |
-
],
|
| 554 |
-
"positive_fold_tail": {
|
| 555 |
-
"null_probability": 0.5,
|
| 556 |
-
"one_sided_probability_at_least_observed": 0.001953125,
|
| 557 |
-
"positive": 9,
|
| 558 |
-
"total": 9
|
| 559 |
-
},
|
| 560 |
-
"repeat_oof_deltas": [
|
| 561 |
-
0.09428967525309362,
|
| 562 |
-
0.10624751083541728,
|
| 563 |
-
0.13687455719435304
|
| 564 |
-
],
|
| 565 |
-
"repeat_oof_t_interval": {
|
| 566 |
-
"confidence": 0.95,
|
| 567 |
-
"critical_t": 4.302652729749462,
|
| 568 |
-
"degrees_of_freedom": 2,
|
| 569 |
-
"interval": [
|
| 570 |
-
0.05790922361228608,
|
| 571 |
-
0.1670319385762899
|
| 572 |
-
],
|
| 573 |
-
"mean": 0.11247058109428798,
|
| 574 |
-
"n": 3,
|
| 575 |
-
"sample_standard_deviation": 0.02196390209122417,
|
| 576 |
-
"standard_error": 0.01268086478482286
|
| 577 |
-
}
|
| 578 |
-
}
|
| 579 |
-
},
|
| 580 |
-
"promotion_gate": {
|
| 581 |
-
"fold_gate_passed": true,
|
| 582 |
-
"observed_positive_fold_deltas": 9,
|
| 583 |
-
"observed_positive_repeat_oof_deltas": 3,
|
| 584 |
-
"passed": true,
|
| 585 |
-
"repeat_gate_passed": true,
|
| 586 |
-
"required_positive_fold_deltas": 7,
|
| 587 |
-
"required_positive_repeat_oof_deltas": 3
|
| 588 |
-
},
|
| 589 |
-
"study": {
|
| 590 |
-
"fold_count": 9,
|
| 591 |
-
"fold_runner_protocol": "malint-qwen35-4l-repeated-matched-cv/1.0",
|
| 592 |
-
"fold_runner_sha256": "c71af81f1182ce860c39c107dd19eff02f1addd8dd646e3eebaadedc4503067c",
|
| 593 |
-
"heldout_split_accessed": false,
|
| 594 |
-
"post_selection_internal_evidence": true,
|
| 595 |
-
"raw_text_written": false,
|
| 596 |
-
"repeat_count": 3,
|
| 597 |
-
"rows_per_repeat": 1100,
|
| 598 |
-
"terminal_evaluation_only": true,
|
| 599 |
-
"threshold": 0.3
|
| 600 |
-
}
|
| 601 |
-
},
|
| 602 |
-
"malint_resources": {
|
| 603 |
-
"all_resource_parity_gates_passed": true,
|
| 604 |
-
"claim_boundary": {
|
| 605 |
-
"development_labels_read": false,
|
| 606 |
-
"heldout_access": false,
|
| 607 |
-
"quality_scored": false,
|
| 608 |
-
"resource_only": true,
|
| 609 |
-
"sota_claim_supported": false,
|
| 610 |
-
"timing_scope": "CUDA model forward with CUDA-resident token IDs and masks; tokenization, loading, disk, and host-to-device transfer excluded"
|
| 611 |
-
},
|
| 612 |
-
"comparisons": {
|
| 613 |
-
"1": {
|
| 614 |
-
"existing_over_mean_kd_throughput_ratio": 1.0456925078172845,
|
| 615 |
-
"mean_kd_minus_existing_parameters": 5,
|
| 616 |
-
"mean_kd_minus_existing_serialized_weight_bytes": -220,
|
| 617 |
-
"mean_kd_over_existing_latency_ratio": 1.0456925078172845,
|
| 618 |
-
"mean_kd_over_existing_peak_allocated_ratio": 1.0007640009038254,
|
| 619 |
-
"paired_round_latency_ratio_summary": {
|
| 620 |
-
"maximum": 1.0527066487704841,
|
| 621 |
-
"mean": 1.0488242542281458,
|
| 622 |
-
"median": 1.0511611345543765,
|
| 623 |
-
"minimum": 1.042604979359577,
|
| 624 |
-
"n": 3,
|
| 625 |
-
"p05": 1.043460594879057,
|
| 626 |
-
"p95": 1.0525520973488733,
|
| 627 |
-
"sample_standard_deviation": 0.005441202851039159
|
| 628 |
-
},
|
| 629 |
-
"paired_round_latency_ratios": [
|
| 630 |
-
1.0511611345543765,
|
| 631 |
-
1.042604979359577,
|
| 632 |
-
1.0527066487704841
|
| 633 |
-
],
|
| 634 |
-
"resource_parity_gate": {
|
| 635 |
-
"four_layers_and_length_256_pass": true,
|
| 636 |
-
"latency_pass": true,
|
| 637 |
-
"latency_ratio": 1.0456925078172845,
|
| 638 |
-
"latency_ratio_limit": 1.05,
|
| 639 |
-
"parameter_pass": true,
|
| 640 |
-
"parameter_relative_difference": 1.4965497464629228e-08,
|
| 641 |
-
"parameter_relative_limit": 0.0001,
|
| 642 |
-
"passed": true,
|
| 643 |
-
"peak_allocated_ratio": 1.0007640009038254,
|
| 644 |
-
"peak_memory_pass": true,
|
| 645 |
-
"peak_memory_ratio_limit": 1.05
|
| 646 |
-
}
|
| 647 |
-
},
|
| 648 |
-
"4": {
|
| 649 |
-
"existing_over_mean_kd_throughput_ratio": 1.0177048568315403,
|
| 650 |
-
"mean_kd_minus_existing_parameters": 5,
|
| 651 |
-
"mean_kd_minus_existing_serialized_weight_bytes": -220,
|
| 652 |
-
"mean_kd_over_existing_latency_ratio": 1.0177048568315403,
|
| 653 |
-
"mean_kd_over_existing_peak_allocated_ratio": 1.0007213257185181,
|
| 654 |
-
"paired_round_latency_ratio_summary": {
|
| 655 |
-
"maximum": 1.0378811577131521,
|
| 656 |
-
"mean": 1.0299336811276214,
|
| 657 |
-
"median": 1.0272510844529572,
|
| 658 |
-
"minimum": 1.0246688012167542,
|
| 659 |
-
"n": 3,
|
| 660 |
-
"p05": 1.0249270295403745,
|
| 661 |
-
"p95": 1.0368181503871325,
|
| 662 |
-
"sample_standard_deviation": 0.007002773360331318
|
| 663 |
-
},
|
| 664 |
-
"paired_round_latency_ratios": [
|
| 665 |
-
1.0272510844529572,
|
| 666 |
-
1.0378811577131521,
|
| 667 |
-
1.0246688012167542
|
| 668 |
-
],
|
| 669 |
-
"resource_parity_gate": {
|
| 670 |
-
"four_layers_and_length_256_pass": true,
|
| 671 |
-
"latency_pass": true,
|
| 672 |
-
"latency_ratio": 1.0177048568315403,
|
| 673 |
-
"latency_ratio_limit": 1.05,
|
| 674 |
-
"parameter_pass": true,
|
| 675 |
-
"parameter_relative_difference": 1.4965497464629228e-08,
|
| 676 |
-
"parameter_relative_limit": 0.0001,
|
| 677 |
-
"passed": true,
|
| 678 |
-
"peak_allocated_ratio": 1.0007213257185181,
|
| 679 |
-
"peak_memory_pass": true,
|
| 680 |
-
"peak_memory_ratio_limit": 1.05
|
| 681 |
-
}
|
| 682 |
-
}
|
| 683 |
-
},
|
| 684 |
-
"rows": [
|
| 685 |
-
{
|
| 686 |
-
"arm": "existing_rightmost",
|
| 687 |
-
"batch": 1,
|
| 688 |
-
"latency_ms_per_example": 18.474629627574572,
|
| 689 |
-
"parameter_count": 334101824,
|
| 690 |
-
"peak_allocated_bytes": 1373150208,
|
| 691 |
-
"worker_count": 3
|
| 692 |
-
},
|
| 693 |
-
{
|
| 694 |
-
"arm": "mean_kd",
|
| 695 |
-
"batch": 1,
|
| 696 |
-
"latency_ms_per_example": 19.318781786253957,
|
| 697 |
-
"parameter_count": 334101829,
|
| 698 |
-
"peak_allocated_bytes": 1374199296,
|
| 699 |
-
"worker_count": 3
|
| 700 |
-
},
|
| 701 |
-
{
|
| 702 |
-
"arm": "existing_rightmost",
|
| 703 |
-
"batch": 4,
|
| 704 |
-
"latency_ms_per_example": 4.589751928502863,
|
| 705 |
-
"parameter_count": 334101824,
|
| 706 |
-
"peak_allocated_bytes": 1454388736,
|
| 707 |
-
"worker_count": 3
|
| 708 |
-
},
|
| 709 |
-
{
|
| 710 |
-
"arm": "mean_kd",
|
| 711 |
-
"batch": 4,
|
| 712 |
-
"latency_ms_per_example": 4.671012829289292,
|
| 713 |
-
"parameter_count": 334101829,
|
| 714 |
-
"peak_allocated_bytes": 1455437824,
|
| 715 |
-
"worker_count": 3
|
| 716 |
-
}
|
| 717 |
-
]
|
| 718 |
-
},
|
| 719 |
-
"privacy": {
|
| 720 |
-
"aggregate_only": true,
|
| 721 |
-
"contains_article_ids": false,
|
| 722 |
-
"contains_logits_probabilities_or_predictions": false,
|
| 723 |
-
"contains_raw_text": false,
|
| 724 |
-
"contains_row_labels": false
|
| 725 |
-
},
|
| 726 |
-
"propaganda_operational_endpoints": [
|
| 727 |
-
{
|
| 728 |
-
"arm": "qwen35-full24l-commonhead",
|
| 729 |
-
"label": "24L",
|
| 730 |
-
"layers": 24,
|
| 731 |
-
"lineage_note": "same five-epoch common-head protocol",
|
| 732 |
-
"macro_f1": {
|
| 733 |
-
"mean": 0.5642638295960175,
|
| 734 |
-
"n": 3,
|
| 735 |
-
"sample_sd": 0.006258972649035587
|
| 736 |
-
},
|
| 737 |
-
"micro_f1": {
|
| 738 |
-
"mean": 0.6264844890231888,
|
| 739 |
-
"n": 3,
|
| 740 |
-
"sample_sd": 0.01398932248989461
|
| 741 |
-
},
|
| 742 |
-
"parameters": 752407360,
|
| 743 |
-
"peak_cuda_allocated_bytes": {
|
| 744 |
-
"mean": 7635339264.0,
|
| 745 |
-
"n": 3,
|
| 746 |
-
"sample_sd": 0.0
|
| 747 |
-
},
|
| 748 |
-
"subset_exact_match": {
|
| 749 |
-
"mean": 0.0303030303030303,
|
| 750 |
-
"n": 3,
|
| 751 |
-
"sample_sd": 0.027773186030035393
|
| 752 |
-
},
|
| 753 |
-
"test_articles": 55,
|
| 754 |
-
"test_seconds_per_article": {
|
| 755 |
-
"mean": 0.586362602903109,
|
| 756 |
-
"n": 3,
|
| 757 |
-
"sample_sd": 0.005979023788731576
|
| 758 |
-
},
|
| 759 |
-
"total_wall_seconds": {
|
| 760 |
-
"mean": 12842.063612925334,
|
| 761 |
-
"n": 3,
|
| 762 |
-
"sample_sd": 94.80225434811064
|
| 763 |
-
}
|
| 764 |
-
},
|
| 765 |
-
{
|
| 766 |
-
"arm": "qwen35-c8q8l-commonhead",
|
| 767 |
-
"label": "8L",
|
| 768 |
-
"layers": 8,
|
| 769 |
-
"lineage_note": "same five-epoch common-head protocol",
|
| 770 |
-
"macro_f1": {
|
| 771 |
-
"mean": 0.5366873350531403,
|
| 772 |
-
"n": 3,
|
| 773 |
-
"sample_sd": 0.017264532146217884
|
| 774 |
-
},
|
| 775 |
-
"micro_f1": {
|
| 776 |
-
"mean": 0.6238830315003392,
|
| 777 |
-
"n": 3,
|
| 778 |
-
"sample_sd": 0.013623163834245318
|
| 779 |
-
},
|
| 780 |
-
"parameters": 420332480,
|
| 781 |
-
"peak_cuda_allocated_bytes": {
|
| 782 |
-
"mean": 4257894400.0,
|
| 783 |
-
"n": 3,
|
| 784 |
-
"sample_sd": 3632373.815194686
|
| 785 |
-
},
|
| 786 |
-
"subset_exact_match": {
|
| 787 |
-
"mean": 0.006060606060606061,
|
| 788 |
-
"n": 3,
|
| 789 |
-
"sample_sd": 0.01049727762162956
|
| 790 |
-
},
|
| 791 |
-
"test_articles": 55,
|
| 792 |
-
"test_seconds_per_article": {
|
| 793 |
-
"mean": 0.206657676709195,
|
| 794 |
-
"n": 3,
|
| 795 |
-
"sample_sd": 0.002071265185964238
|
| 796 |
-
},
|
| 797 |
-
"total_wall_seconds": {
|
| 798 |
-
"mean": 4291.582853937677,
|
| 799 |
-
"n": 3,
|
| 800 |
-
"sample_sd": 44.49609583978631
|
| 801 |
-
}
|
| 802 |
-
},
|
| 803 |
-
{
|
| 804 |
-
"arm": "qwen35-standalone4l-commonhead",
|
| 805 |
-
"label": "4L",
|
| 806 |
-
"layers": 4,
|
| 807 |
-
"lineage_note": "same five-epoch common-head protocol",
|
| 808 |
-
"macro_f1": {
|
| 809 |
-
"mean": 0.5875679559555907,
|
| 810 |
-
"n": 3,
|
| 811 |
-
"sample_sd": 0.007165753447433348
|
| 812 |
-
},
|
| 813 |
-
"micro_f1": {
|
| 814 |
-
"mean": 0.647978512363484,
|
| 815 |
-
"n": 3,
|
| 816 |
-
"sample_sd": 0.0018739735696546695
|
| 817 |
-
},
|
| 818 |
-
"parameters": 210903360,
|
| 819 |
-
"peak_cuda_allocated_bytes": {
|
| 820 |
-
"mean": 2143655936.0,
|
| 821 |
-
"n": 3,
|
| 822 |
-
"sample_sd": 0.0
|
| 823 |
-
},
|
| 824 |
-
"subset_exact_match": {
|
| 825 |
-
"mean": 0.006060606060606061,
|
| 826 |
-
"n": 3,
|
| 827 |
-
"sample_sd": 0.01049727762162956
|
| 828 |
-
},
|
| 829 |
-
"test_articles": 55,
|
| 830 |
-
"test_seconds_per_article": {
|
| 831 |
-
"mean": 0.08543555729062946,
|
| 832 |
-
"n": 3,
|
| 833 |
-
"sample_sd": 0.0005980667826383295
|
| 834 |
-
},
|
| 835 |
-
"total_wall_seconds": {
|
| 836 |
-
"mean": 1643.6388439506602,
|
| 837 |
-
"n": 3,
|
| 838 |
-
"sample_sd": 9.226529522850198
|
| 839 |
-
}
|
| 840 |
-
}
|
| 841 |
-
],
|
| 842 |
-
"schema": "qwen35-4l-cross-task-base-card-figures-v1",
|
| 843 |
-
"status": "complete",
|
| 844 |
-
"task_free_compression_ladder": [
|
| 845 |
-
{
|
| 846 |
-
"layers": 24,
|
| 847 |
-
"name": "24L upstream text teacher",
|
| 848 |
-
"parameters": 752393024,
|
| 849 |
-
"weight_bytes": 1504825632,
|
| 850 |
-
"weight_mib": 1435.1135559082031
|
| 851 |
-
},
|
| 852 |
-
{
|
| 853 |
-
"layers": 8,
|
| 854 |
-
"name": "8L task-free KD",
|
| 855 |
-
"parameters": 420318144,
|
| 856 |
-
"weight_bytes": 840647880,
|
| 857 |
-
"weight_mib": 801.7042922973633
|
| 858 |
-
},
|
| 859 |
-
{
|
| 860 |
-
"layers": 6,
|
| 861 |
-
"name": "6L task-free KD",
|
| 862 |
-
"parameters": 377207424,
|
| 863 |
-
"weight_bytes": 754423424,
|
| 864 |
-
"weight_mib": 719.4742431640625
|
| 865 |
-
},
|
| 866 |
-
{
|
| 867 |
-
"layers": 4,
|
| 868 |
-
"name": "4L task-free KD root",
|
| 869 |
-
"parameters": 334096704,
|
| 870 |
-
"weight_bytes": 668198976,
|
| 871 |
-
"weight_mib": 637.2442016601562
|
| 872 |
-
}
|
| 873 |
-
],
|
| 874 |
-
"uno_q_controlled_depth": {
|
| 875 |
-
"boundary": "Matched V128k/F16 runtime comparison. The 4L arm is the historical early-path proxy, not the final common-head quality model.",
|
| 876 |
-
"claim_boundaries": [
|
| 877 |
-
"Only layer_effect_primary is used for same-vocabulary F16 depth-effect estimates.",
|
| 878 |
-
"Only vocabulary_effect_primary is used for exact-remap vocabulary-effect estimates.",
|
| 879 |
-
"Operational and cross-generation groups are descriptive because precision, heads, layer maps, training lineages, or architectures differ.",
|
| 880 |
-
"llama-bench embedding-mode latency measures backbone/classifier compute but does not validate evidence-span quality.",
|
| 881 |
-
"CPU occupancy is instantaneous saturation; core-seconds is a compute-duration proxy, not measured energy.",
|
| 882 |
-
"No power meter was used, so energy consumption is not reported.",
|
| 883 |
-
"Temperature is device- and ambient-specific and is descriptive, not a universal hardware constant."
|
| 884 |
-
],
|
| 885 |
-
"rows": [
|
| 886 |
-
{
|
| 887 |
-
"layers": 8,
|
| 888 |
-
"metric": "artifact_mib",
|
| 889 |
-
"metric_label": "Artifact size",
|
| 890 |
-
"model": "8L",
|
| 891 |
-
"normalized_8l_percent": 100.0
|
| 892 |
-
},
|
| 893 |
-
{
|
| 894 |
-
"layers": 6,
|
| 895 |
-
"metric": "artifact_mib",
|
| 896 |
-
"metric_label": "Artifact size",
|
| 897 |
-
"model": "6L",
|
| 898 |
-
"normalized_8l_percent": 85.59176355424236
|
| 899 |
-
},
|
| 900 |
-
{
|
| 901 |
-
"layers": 4,
|
| 902 |
-
"metric": "artifact_mib",
|
| 903 |
-
"metric_label": "Artifact size",
|
| 904 |
-
"model": "4L proxy",
|
| 905 |
-
"normalized_8l_percent": 72.2607217304372
|
| 906 |
-
},
|
| 907 |
-
{
|
| 908 |
-
"layers": 8,
|
| 909 |
-
"metric": "latency_pp128_ms",
|
| 910 |
-
"metric_label": "PP128 latency",
|
| 911 |
-
"model": "8L",
|
| 912 |
-
"normalized_8l_percent": 100.0
|
| 913 |
-
},
|
| 914 |
-
{
|
| 915 |
-
"layers": 6,
|
| 916 |
-
"metric": "latency_pp128_ms",
|
| 917 |
-
"metric_label": "PP128 latency",
|
| 918 |
-
"model": "6L",
|
| 919 |
-
"normalized_8l_percent": 76.36565126547553
|
| 920 |
-
},
|
| 921 |
-
{
|
| 922 |
-
"layers": 4,
|
| 923 |
-
"metric": "latency_pp128_ms",
|
| 924 |
-
"metric_label": "PP128 latency",
|
| 925 |
-
"model": "4L proxy",
|
| 926 |
-
"normalized_8l_percent": 50.16216836129361
|
| 927 |
-
},
|
| 928 |
-
{
|
| 929 |
-
"layers": 8,
|
| 930 |
-
"metric": "peak_rss_mib",
|
| 931 |
-
"metric_label": "Peak RSS",
|
| 932 |
-
"model": "8L",
|
| 933 |
-
"normalized_8l_percent": 100.0
|
| 934 |
-
},
|
| 935 |
-
{
|
| 936 |
-
"layers": 6,
|
| 937 |
-
"metric": "peak_rss_mib",
|
| 938 |
-
"metric_label": "Peak RSS",
|
| 939 |
-
"model": "6L",
|
| 940 |
-
"normalized_8l_percent": 87.50983221571457
|
| 941 |
-
},
|
| 942 |
-
{
|
| 943 |
-
"layers": 4,
|
| 944 |
-
"metric": "peak_rss_mib",
|
| 945 |
-
"metric_label": "Peak RSS",
|
| 946 |
-
"model": "4L proxy",
|
| 947 |
-
"normalized_8l_percent": 78.19200407435702
|
| 948 |
-
},
|
| 949 |
-
{
|
| 950 |
-
"layers": 8,
|
| 951 |
-
"metric": "mean_core_seconds_per_round",
|
| 952 |
-
"metric_label": "Core-seconds",
|
| 953 |
-
"model": "8L",
|
| 954 |
-
"normalized_8l_percent": 100.0
|
| 955 |
-
},
|
| 956 |
-
{
|
| 957 |
-
"layers": 6,
|
| 958 |
-
"metric": "mean_core_seconds_per_round",
|
| 959 |
-
"metric_label": "Core-seconds",
|
| 960 |
-
"model": "6L",
|
| 961 |
-
"normalized_8l_percent": 76.16702179915283
|
| 962 |
-
},
|
| 963 |
-
{
|
| 964 |
-
"layers": 4,
|
| 965 |
-
"metric": "mean_core_seconds_per_round",
|
| 966 |
-
"metric_label": "Core-seconds",
|
| 967 |
-
"model": "4L proxy",
|
| 968 |
-
"normalized_8l_percent": 51.03561371028788
|
| 969 |
-
}
|
| 970 |
-
]
|
| 971 |
-
}
|
| 972 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@@ -0,0 +1,92 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""Figure 10: the published root against every arm measured under the frozen protocol, on every metric."""
|
| 2 |
+
import csv, json, pathlib
|
| 3 |
+
import matplotlib
|
| 4 |
+
matplotlib.use("Agg")
|
| 5 |
+
import matplotlib.pyplot as plt
|
| 6 |
+
import numpy as np
|
| 7 |
+
|
| 8 |
+
S = pathlib.Path("/tmp/claude-1000/-home-mp-ubuntu-Projects-paper-qwen-intent-classifier/"
|
| 9 |
+
"0af32b5a-ac18-4e73-a3ad-1c863d0154af/scratchpad")
|
| 10 |
+
OUT = S / "fig2"
|
| 11 |
+
SURFACE = "#fcfcfb"; INK, INK2, MUTED = "#1a1a19", "#4a4a47", "#8a8a85"; GRID = "#e5e5e1"
|
| 12 |
+
C = {"blue": "#2a78d6", "orange": "#eb6834", "aqua": "#1baf7a", "yellow": "#eda100", "violet": "#4a3aa7"}
|
| 13 |
+
plt.rcParams.update({
|
| 14 |
+
"figure.facecolor": SURFACE, "axes.facecolor": SURFACE, "savefig.facecolor": SURFACE,
|
| 15 |
+
"axes.edgecolor": GRID, "axes.linewidth": 0.8, "axes.labelcolor": INK2,
|
| 16 |
+
"xtick.color": MUTED, "ytick.color": MUTED, "text.color": INK,
|
| 17 |
+
"font.size": 9, "axes.titlesize": 10.5, "axes.titleweight": "semibold",
|
| 18 |
+
"grid.color": GRID, "grid.linewidth": 0.7,
|
| 19 |
+
"xtick.major.size": 0, "ytick.major.size": 0,
|
| 20 |
+
"legend.frameon": False, "figure.dpi": 160,
|
| 21 |
+
})
|
| 22 |
+
|
| 23 |
+
arms = json.loads((S / "merged_arms.json").read_text())
|
| 24 |
+
MI = 1024 ** 2
|
| 25 |
+
rows = {r["arm_id"]: r for r in csv.DictReader(open(str(S / "hf/bench/61_arm_bf16.csv")))}
|
| 26 |
+
g = json.loads(pathlib.Path("/home/mp_ubuntu/Projects/paper_qwen_intent_classifier/agents/sessions/"
|
| 27 |
+
"2026-09-20/S-20260920-taskblind-ladder-v1/MEASUREMENTS.json").read_text())
|
| 28 |
+
gridN = {"15,380": "8192", "23,551": "16384", "39,866": "32768", "72,455": "65536"}
|
| 29 |
+
for a in arms:
|
| 30 |
+
if a["new"]:
|
| 31 |
+
n = next(v for k, v in gridN.items() if k in a["arm_id"])
|
| 32 |
+
a["peak_rss_mib"] = g["arms"][n]["BF16"]["peak_tree_rss_bytes"] / MI
|
| 33 |
+
else:
|
| 34 |
+
a["peak_rss_mib"] = float(rows[a["arm_id"]]["peak_rss_mib"])
|
| 35 |
+
|
| 36 |
+
FINAL = "39,866"
|
| 37 |
+
def fam(a):
|
| 38 |
+
if a["new"]: return "final" if FINAL in a["arm_id"] else "grid"
|
| 39 |
+
return "qwen" if a["arm_id"].startswith(("qwen35", "qwen25")) else "encoder"
|
| 40 |
+
COL = {"encoder": MUTED, "qwen": C["violet"], "grid": C["blue"], "final": C["orange"]}
|
| 41 |
+
|
| 42 |
+
PANELS = [
|
| 43 |
+
("macro_f1_mean", "macro F1 — higher is better", True, False, "{:.4f}"),
|
| 44 |
+
("p50_ms", "document p50, ms — lower is better", False, True, "{:.1f}"),
|
| 45 |
+
("peak_rss_mib", "peak process-tree RSS, MiB — lower is better", False, False, "{:.0f}"),
|
| 46 |
+
("weight_file_mib", "BF16 weight file, MiB — lower is better", False, True, "{:.0f}"),
|
| 47 |
+
("gpu_j_per_doc", "GPU joules per document — lower is better", False, True, "{:.2f}"),
|
| 48 |
+
]
|
| 49 |
+
|
| 50 |
+
fig, axes = plt.subplots(len(PANELS), 1, figsize=(11.0, 7.4))
|
| 51 |
+
for ax, (key, title, higher, logx, fmt) in zip(axes, PANELS):
|
| 52 |
+
ax.set_axisbelow(True)
|
| 53 |
+
ax.grid(True, axis="x", alpha=0.9)
|
| 54 |
+
for s in ("top", "right", "left"): ax.spines[s].set_visible(False)
|
| 55 |
+
ax.spines["bottom"].set_color(GRID)
|
| 56 |
+
vals = [a[key] for a in arms]
|
| 57 |
+
root = next(a for a in arms if fam(a) == "final")
|
| 58 |
+
order = sorted(vals, reverse=higher)
|
| 59 |
+
rank = order.index(root[key]) + 1
|
| 60 |
+
for a in arms:
|
| 61 |
+
f = fam(a)
|
| 62 |
+
if f == "final": continue
|
| 63 |
+
ax.plot([a[key], a[key]], [0.18, 0.82], lw=1.6 if f == "grid" else 1.1,
|
| 64 |
+
color=COL[f], alpha=0.95 if f == "grid" else 0.45, solid_capstyle="butt", zorder=3)
|
| 65 |
+
ax.plot([root[key]], [0.5], marker="*", ms=17, color=C["orange"],
|
| 66 |
+
markeredgecolor=SURFACE, markeredgewidth=1.2, zorder=6)
|
| 67 |
+
if logx: ax.set_xscale("log")
|
| 68 |
+
ax.set_yticks([]); ax.set_ylim(0, 1.25)
|
| 69 |
+
ax.set_title(title, loc="left", pad=6)
|
| 70 |
+
good = "top" if rank <= len(arms) / 3 else ("middle" if rank <= 2 * len(arms) / 3 else "bottom")
|
| 71 |
+
ax.annotate(f"this root {fmt.format(root[key])} · rank {rank} of {len(arms)} ({good} third)",
|
| 72 |
+
(0.998, 1.06), xycoords="axes fraction", ha="right", va="bottom",
|
| 73 |
+
fontsize=8.5, color=C["orange"], fontweight="semibold")
|
| 74 |
+
lo, hi = min(vals), max(vals)
|
| 75 |
+
ax.annotate(f"best {fmt.format(order[0])}", (hi if higher else lo, 0.02),
|
| 76 |
+
xycoords=("data", "axes fraction"), ha="right" if higher else "left",
|
| 77 |
+
va="bottom", fontsize=7, color=MUTED)
|
| 78 |
+
|
| 79 |
+
handles = [plt.Line2D([], [], color=COL["encoder"], lw=2, label="encoder families (46)"),
|
| 80 |
+
plt.Line2D([], [], color=COL["qwen"], lw=2, label="earlier Qwen arms (15)"),
|
| 81 |
+
plt.Line2D([], [], color=COL["grid"], lw=2, label="task-blind vocabulary grid (4)"),
|
| 82 |
+
plt.Line2D([], [], color=C["orange"], marker="*", ms=13, lw=0, label="this root: 4L, 39,866 EN/KO")]
|
| 83 |
+
fig.legend(handles=handles, loc="lower center", ncol=4, fontsize=8.5, bbox_to_anchor=(0.5, 0.062))
|
| 84 |
+
fig.suptitle("The published root against all 65 arms, on every metric the protocol measured",
|
| 85 |
+
fontsize=12, fontweight="semibold", y=0.988)
|
| 86 |
+
fig.text(0.5, 0.030, "Each tick is one arm. Quality is a transfer probe with a fresh 14-label head, three seeds; the other four are single measured values.",
|
| 87 |
+
ha="center", fontsize=7.5, color=MUTED)
|
| 88 |
+
fig.text(0.5, 0.008, "The four grid arms were measured in a later session; a same-host check on an earlier cut put the host effect at 1.05x on BF16 latency and 0.99x on RSS, so read small latency gaps with that in mind.",
|
| 89 |
+
ha="center", fontsize=7.5, color=MUTED)
|
| 90 |
+
fig.subplots_adjust(left=0.035, right=0.985, top=0.925, bottom=0.135, hspace=0.72)
|
| 91 |
+
fig.savefig(OUT / "10_root_on_every_metric.png")
|
| 92 |
+
print("wrote 10_root_on_every_metric.png")
|
|
@@ -1,14 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"schema_version": "standalone4l-v128k-figures-v1",
|
| 3 |
-
"quality_scope": "existing opened SemEval transfer evidence; 128k vs 248k is not causal because lineages differ",
|
| 4 |
-
"source_files": [
|
| 5 |
-
"results/standalone4l_base/20260731-v1/huggingface_release_v3/benchmark/arm_summary.csv",
|
| 6 |
-
"results/standalone4l_base/20260731-v1/huggingface_release_v3/benchmark/compression_ladder.csv",
|
| 7 |
-
"results/standalone4l_base/20260731-v1/huggingface_release_v3/benchmark/resource_metrics.csv"
|
| 8 |
-
],
|
| 9 |
-
"figures": [
|
| 10 |
-
"v128k_vocabulary_size.png",
|
| 11 |
-
"v128k_quality_storage.png"
|
| 12 |
-
],
|
| 13 |
-
"new_headless_quality": "NR: no new downstream benchmark run"
|
| 14 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@@ -1,891 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"schema": "standalone4l-complete-transfer-benchmark-v2",
|
| 3 |
-
"status": "complete_from_frozen_reports",
|
| 4 |
-
"claim_scope": "Exploratory fresh-head transfer evidence on one English task and a previously opened 55-article SemEval-derived test split.",
|
| 5 |
-
"headless_root_requires_adaptation": true,
|
| 6 |
-
"test_previously_opened": true,
|
| 7 |
-
"test_articles": 55,
|
| 8 |
-
"test_windows": 434,
|
| 9 |
-
"seeds": [
|
| 10 |
-
41,
|
| 11 |
-
42,
|
| 12 |
-
43
|
| 13 |
-
],
|
| 14 |
-
"three_seed_sd_is_not_confidence_interval": true,
|
| 15 |
-
"multilingual_downstream_evaluation_performed": false,
|
| 16 |
-
"upstream_multilingual_metadata": {
|
| 17 |
-
"source": "Qwen/Qwen3.5-0.8B official model card",
|
| 18 |
-
"languages_and_dialects_claimed_upstream": 201,
|
| 19 |
-
"scope": "inherited tokenizer and architecture coverage only"
|
| 20 |
-
},
|
| 21 |
-
"protocol": {
|
| 22 |
-
"epochs": 5,
|
| 23 |
-
"primary_epoch": 5,
|
| 24 |
-
"early_stopping": false,
|
| 25 |
-
"max_length": 256,
|
| 26 |
-
"stride": 128,
|
| 27 |
-
"window_target": "span-aligned technique labels",
|
| 28 |
-
"article_aggregation": "labelwise maximum probability",
|
| 29 |
-
"micro_batch_size": 1,
|
| 30 |
-
"gradient_accumulation": 32,
|
| 31 |
-
"effective_batch_size": 32,
|
| 32 |
-
"learning_rate": 2e-05,
|
| 33 |
-
"weight_decay": 0.01,
|
| 34 |
-
"gradient_clip": 1.0,
|
| 35 |
-
"optimizer": "AdamW",
|
| 36 |
-
"dtype": "bfloat16",
|
| 37 |
-
"thresholds": [
|
| 38 |
-
0.05,
|
| 39 |
-
0.1,
|
| 40 |
-
0.15,
|
| 41 |
-
0.2,
|
| 42 |
-
0.25,
|
| 43 |
-
0.3,
|
| 44 |
-
0.35,
|
| 45 |
-
0.4,
|
| 46 |
-
0.45,
|
| 47 |
-
0.5,
|
| 48 |
-
0.55,
|
| 49 |
-
0.6,
|
| 50 |
-
0.65,
|
| 51 |
-
0.7,
|
| 52 |
-
0.75,
|
| 53 |
-
0.8,
|
| 54 |
-
0.85,
|
| 55 |
-
0.9
|
| 56 |
-
],
|
| 57 |
-
"threshold_selection": "epoch-5 calibration macro-F1, then micro-F1"
|
| 58 |
-
},
|
| 59 |
-
"arms": {
|
| 60 |
-
"task_agnostic_base": {
|
| 61 |
-
"display": "Task-agnostic KD base",
|
| 62 |
-
"lineage": "root base plus a fresh 14-label transfer head",
|
| 63 |
-
"metrics": {
|
| 64 |
-
"macro_f1": {
|
| 65 |
-
"values": [
|
| 66 |
-
0.5883093293517536,
|
| 67 |
-
0.6028379036218096,
|
| 68 |
-
0.5712816197954101
|
| 69 |
-
],
|
| 70 |
-
"mean": 0.5874762842563245,
|
| 71 |
-
"sample_sd": 0.01579462678670608,
|
| 72 |
-
"n": 3
|
| 73 |
-
},
|
| 74 |
-
"micro_f1": {
|
| 75 |
-
"values": [
|
| 76 |
-
0.6468085106382978,
|
| 77 |
-
0.6655172413793103,
|
| 78 |
-
0.6465256797583082
|
| 79 |
-
],
|
| 80 |
-
"mean": 0.6529504772586388,
|
| 81 |
-
"sample_sd": 0.010884055708931178,
|
| 82 |
-
"n": 3
|
| 83 |
-
},
|
| 84 |
-
"exact_match": {
|
| 85 |
-
"values": [
|
| 86 |
-
0.05454545454545454,
|
| 87 |
-
0.03636363636363636,
|
| 88 |
-
0.0
|
| 89 |
-
],
|
| 90 |
-
"mean": 0.0303030303030303,
|
| 91 |
-
"sample_sd": 0.027773186030035393,
|
| 92 |
-
"n": 3
|
| 93 |
-
},
|
| 94 |
-
"predicted_positive_rate": {
|
| 95 |
-
"values": [
|
| 96 |
-
0.5714285714285714,
|
| 97 |
-
0.4090909090909091,
|
| 98 |
-
0.5155844155844156
|
| 99 |
-
],
|
| 100 |
-
"mean": 0.4987012987012987,
|
| 101 |
-
"sample_sd": 0.08247520160679181,
|
| 102 |
-
"n": 3
|
| 103 |
-
}
|
| 104 |
-
},
|
| 105 |
-
"resources": {
|
| 106 |
-
"parameters": {
|
| 107 |
-
"values": [
|
| 108 |
-
334111040.0,
|
| 109 |
-
334111040.0,
|
| 110 |
-
334111040.0
|
| 111 |
-
],
|
| 112 |
-
"mean": 334111040.0,
|
| 113 |
-
"sample_sd": 0.0,
|
| 114 |
-
"n": 3
|
| 115 |
-
},
|
| 116 |
-
"peak_cuda_allocated_bytes": {
|
| 117 |
-
"values": [
|
| 118 |
-
3380975616.0,
|
| 119 |
-
3380975616.0,
|
| 120 |
-
3380975616.0
|
| 121 |
-
],
|
| 122 |
-
"mean": 3380975616.0,
|
| 123 |
-
"sample_sd": 0.0,
|
| 124 |
-
"n": 3
|
| 125 |
-
},
|
| 126 |
-
"peak_cuda_reserved_bytes": {
|
| 127 |
-
"values": [
|
| 128 |
-
3519021056.0,
|
| 129 |
-
3519021056.0,
|
| 130 |
-
3510632448.0
|
| 131 |
-
],
|
| 132 |
-
"mean": 3516224853.3333335,
|
| 133 |
-
"sample_sd": 4843165.086926248,
|
| 134 |
-
"n": 3
|
| 135 |
-
},
|
| 136 |
-
"total_wall_seconds": {
|
| 137 |
-
"values": [
|
| 138 |
-
1583.0297977919981,
|
| 139 |
-
1588.617994428001,
|
| 140 |
-
1603.4310500120046
|
| 141 |
-
],
|
| 142 |
-
"mean": 1591.692947410668,
|
| 143 |
-
"sample_sd": 10.54249851412311,
|
| 144 |
-
"n": 3
|
| 145 |
-
},
|
| 146 |
-
"test_seconds_per_article": {
|
| 147 |
-
"values": [
|
| 148 |
-
0.08164824154545468,
|
| 149 |
-
0.07871871667289683,
|
| 150 |
-
0.08355205721825107
|
| 151 |
-
],
|
| 152 |
-
"mean": 0.08130633847886752,
|
| 153 |
-
"sample_sd": 0.0024347419754560395,
|
| 154 |
-
"n": 3
|
| 155 |
-
}
|
| 156 |
-
},
|
| 157 |
-
"thresholds": [
|
| 158 |
-
0.15,
|
| 159 |
-
0.3,
|
| 160 |
-
0.1
|
| 161 |
-
]
|
| 162 |
-
},
|
| 163 |
-
"structural_copy_control": {
|
| 164 |
-
"display": "Structural copy, no task-free KD",
|
| 165 |
-
"lineage": "same four-layer structure without task-free KD",
|
| 166 |
-
"metrics": {
|
| 167 |
-
"macro_f1": {
|
| 168 |
-
"values": [
|
| 169 |
-
0.5169424579262757,
|
| 170 |
-
0.5349405779540175,
|
| 171 |
-
0.5023778364496152
|
| 172 |
-
],
|
| 173 |
-
"mean": 0.5180869574433028,
|
| 174 |
-
"sample_sd": 0.016311512588628545,
|
| 175 |
-
"n": 3
|
| 176 |
-
},
|
| 177 |
-
"micro_f1": {
|
| 178 |
-
"values": [
|
| 179 |
-
0.5773955773955775,
|
| 180 |
-
0.6083445491251682,
|
| 181 |
-
0.585635359116022
|
| 182 |
-
],
|
| 183 |
-
"mean": 0.5904584952122559,
|
| 184 |
-
"sample_sd": 0.01602830914762526,
|
| 185 |
-
"n": 3
|
| 186 |
-
},
|
| 187 |
-
"exact_match": {
|
| 188 |
-
"values": [
|
| 189 |
-
0.0,
|
| 190 |
-
0.0,
|
| 191 |
-
0.0
|
| 192 |
-
],
|
| 193 |
-
"mean": 0.0,
|
| 194 |
-
"sample_sd": 0.0,
|
| 195 |
-
"n": 3
|
| 196 |
-
},
|
| 197 |
-
"predicted_positive_rate": {
|
| 198 |
-
"values": [
|
| 199 |
-
0.712987012987013,
|
| 200 |
-
0.6207792207792208,
|
| 201 |
-
0.5961038961038961
|
| 202 |
-
],
|
| 203 |
-
"mean": 0.6432900432900432,
|
| 204 |
-
"sample_sd": 0.06160737456395461,
|
| 205 |
-
"n": 3
|
| 206 |
-
}
|
| 207 |
-
},
|
| 208 |
-
"resources": {
|
| 209 |
-
"parameters": {
|
| 210 |
-
"values": [
|
| 211 |
-
334111040.0,
|
| 212 |
-
334111040.0,
|
| 213 |
-
334111040.0
|
| 214 |
-
],
|
| 215 |
-
"mean": 334111040.0,
|
| 216 |
-
"sample_sd": 0.0,
|
| 217 |
-
"n": 3
|
| 218 |
-
},
|
| 219 |
-
"peak_cuda_allocated_bytes": {
|
| 220 |
-
"values": [
|
| 221 |
-
3380975616.0,
|
| 222 |
-
3380975616.0,
|
| 223 |
-
3380975616.0
|
| 224 |
-
],
|
| 225 |
-
"mean": 3380975616.0,
|
| 226 |
-
"sample_sd": 0.0,
|
| 227 |
-
"n": 3
|
| 228 |
-
},
|
| 229 |
-
"peak_cuda_reserved_bytes": {
|
| 230 |
-
"values": [
|
| 231 |
-
3519021056.0,
|
| 232 |
-
3519021056.0,
|
| 233 |
-
3510632448.0
|
| 234 |
-
],
|
| 235 |
-
"mean": 3516224853.3333335,
|
| 236 |
-
"sample_sd": 4843165.086926248,
|
| 237 |
-
"n": 3
|
| 238 |
-
},
|
| 239 |
-
"total_wall_seconds": {
|
| 240 |
-
"values": [
|
| 241 |
-
1575.5905925639963,
|
| 242 |
-
1623.5639530190092,
|
| 243 |
-
1629.195114912989
|
| 244 |
-
],
|
| 245 |
-
"mean": 1609.4498868319981,
|
| 246 |
-
"sample_sd": 29.457874198835384,
|
| 247 |
-
"n": 3
|
| 248 |
-
},
|
| 249 |
-
"test_seconds_per_article": {
|
| 250 |
-
"values": [
|
| 251 |
-
0.07877608058201573,
|
| 252 |
-
0.08307423260041767,
|
| 253 |
-
0.08163602005436339
|
| 254 |
-
],
|
| 255 |
-
"mean": 0.08116211107893226,
|
| 256 |
-
"sample_sd": 0.002187914527825257,
|
| 257 |
-
"n": 3
|
| 258 |
-
}
|
| 259 |
-
},
|
| 260 |
-
"thresholds": [
|
| 261 |
-
0.05,
|
| 262 |
-
0.2,
|
| 263 |
-
0.2
|
| 264 |
-
]
|
| 265 |
-
},
|
| 266 |
-
"existing_specialized_separate_lineage": {
|
| 267 |
-
"display": "Existing specialized 4L, separate lineage",
|
| 268 |
-
"lineage": "historical task-specific shrink/distillation lineage",
|
| 269 |
-
"metrics": {
|
| 270 |
-
"macro_f1": {
|
| 271 |
-
"values": [
|
| 272 |
-
0.5832848452288824,
|
| 273 |
-
0.5958405181160762,
|
| 274 |
-
0.5835785045218137
|
| 275 |
-
],
|
| 276 |
-
"mean": 0.5875679559555907,
|
| 277 |
-
"sample_sd": 0.007165753447433348,
|
| 278 |
-
"n": 3
|
| 279 |
-
},
|
| 280 |
-
"micro_f1": {
|
| 281 |
-
"values": [
|
| 282 |
-
0.6494252873563218,
|
| 283 |
-
0.6458616010854816,
|
| 284 |
-
0.6486486486486486
|
| 285 |
-
],
|
| 286 |
-
"mean": 0.647978512363484,
|
| 287 |
-
"sample_sd": 0.0018739735696546695,
|
| 288 |
-
"n": 3
|
| 289 |
-
},
|
| 290 |
-
"exact_match": {
|
| 291 |
-
"values": [
|
| 292 |
-
0.0,
|
| 293 |
-
0.0,
|
| 294 |
-
0.01818181818181818
|
| 295 |
-
],
|
| 296 |
-
"mean": 0.006060606060606061,
|
| 297 |
-
"sample_sd": 0.01049727762162956,
|
| 298 |
-
"n": 3
|
| 299 |
-
},
|
| 300 |
-
"predicted_positive_rate": {
|
| 301 |
-
"values": [
|
| 302 |
-
0.5597402597402598,
|
| 303 |
-
0.612987012987013,
|
| 304 |
-
0.5688311688311688
|
| 305 |
-
],
|
| 306 |
-
"mean": 0.5805194805194805,
|
| 307 |
-
"sample_sd": 0.02848274311618351,
|
| 308 |
-
"n": 3
|
| 309 |
-
}
|
| 310 |
-
},
|
| 311 |
-
"resources": {
|
| 312 |
-
"parameters": {
|
| 313 |
-
"values": [
|
| 314 |
-
210903360.0,
|
| 315 |
-
210903360.0,
|
| 316 |
-
210903360.0
|
| 317 |
-
],
|
| 318 |
-
"mean": 210903360.0,
|
| 319 |
-
"sample_sd": 0.0,
|
| 320 |
-
"n": 3
|
| 321 |
-
},
|
| 322 |
-
"peak_cuda_allocated_bytes": {
|
| 323 |
-
"values": [
|
| 324 |
-
2143655936.0,
|
| 325 |
-
2143655936.0,
|
| 326 |
-
2143655936.0
|
| 327 |
-
],
|
| 328 |
-
"mean": 2143655936.0,
|
| 329 |
-
"sample_sd": 0.0,
|
| 330 |
-
"n": 3
|
| 331 |
-
},
|
| 332 |
-
"peak_cuda_reserved_bytes": {
|
| 333 |
-
"values": [
|
| 334 |
-
2281701376.0,
|
| 335 |
-
2281701376.0,
|
| 336 |
-
2273312768.0
|
| 337 |
-
],
|
| 338 |
-
"mean": 2278905173.3333335,
|
| 339 |
-
"sample_sd": 4843165.086926248,
|
| 340 |
-
"n": 3
|
| 341 |
-
},
|
| 342 |
-
"total_wall_seconds": {
|
| 343 |
-
"values": [
|
| 344 |
-
1634.488557530014,
|
| 345 |
-
1652.9397718419787,
|
| 346 |
-
1643.488202479988
|
| 347 |
-
],
|
| 348 |
-
"mean": 1643.6388439506602,
|
| 349 |
-
"sample_sd": 9.226529522850198,
|
| 350 |
-
"n": 3
|
| 351 |
-
},
|
| 352 |
-
"test_seconds_per_article": {
|
| 353 |
-
"values": [
|
| 354 |
-
0.0847959352360869,
|
| 355 |
-
0.08598086021798239,
|
| 356 |
-
0.08552987641781908
|
| 357 |
-
],
|
| 358 |
-
"mean": 0.08543555729062946,
|
| 359 |
-
"sample_sd": 0.0005980667826383295,
|
| 360 |
-
"n": 3
|
| 361 |
-
}
|
| 362 |
-
},
|
| 363 |
-
"thresholds": [
|
| 364 |
-
0.1,
|
| 365 |
-
0.05,
|
| 366 |
-
0.05
|
| 367 |
-
]
|
| 368 |
-
}
|
| 369 |
-
},
|
| 370 |
-
"same_seed_macro_f1_deltas_base_minus_control": [
|
| 371 |
-
0.07136687142547793,
|
| 372 |
-
0.06789732566779205,
|
| 373 |
-
0.06890378334579494
|
| 374 |
-
],
|
| 375 |
-
"mean_macro_f1_delta_base_minus_control": 0.06938932681302164,
|
| 376 |
-
"publication_gate_passed": true,
|
| 377 |
-
"per_label_summary": [
|
| 378 |
-
{
|
| 379 |
-
"arm": "task_agnostic_base",
|
| 380 |
-
"label": "Appeal_to_Authority",
|
| 381 |
-
"support": 11,
|
| 382 |
-
"precision_mean": 0.3958614484930274,
|
| 383 |
-
"precision_sample_sd": 0.1238064457010119,
|
| 384 |
-
"recall_mean": 0.5757575757575757,
|
| 385 |
-
"recall_sample_sd": 0.05248638810814781,
|
| 386 |
-
"f1_mean": 0.4657088122605364,
|
| 387 |
-
"f1_sample_sd": 0.1020990126979069
|
| 388 |
-
},
|
| 389 |
-
{
|
| 390 |
-
"arm": "task_agnostic_base",
|
| 391 |
-
"label": "Appeal_to_fear-prejudice",
|
| 392 |
-
"support": 24,
|
| 393 |
-
"precision_mean": 0.5540244550770866,
|
| 394 |
-
"precision_sample_sd": 0.02525673990402998,
|
| 395 |
-
"recall_mean": 0.7361111111111112,
|
| 396 |
-
"recall_sample_sd": 0.13393959390267993,
|
| 397 |
-
"f1_mean": 0.6277521761392729,
|
| 398 |
-
"f1_sample_sd": 0.04994879016183189
|
| 399 |
-
},
|
| 400 |
-
{
|
| 401 |
-
"arm": "task_agnostic_base",
|
| 402 |
-
"label": "Bandwagon,Reductio_ad_hitlerum",
|
| 403 |
-
"support": 7,
|
| 404 |
-
"precision_mean": 0.4851851851851852,
|
| 405 |
-
"precision_sample_sd": 0.07882887649552875,
|
| 406 |
-
"recall_mean": 0.8095238095238095,
|
| 407 |
-
"recall_sample_sd": 0.08247860988423221,
|
| 408 |
-
"f1_mean": 0.6006778309409889,
|
| 409 |
-
"f1_sample_sd": 0.04793776258918466
|
| 410 |
-
},
|
| 411 |
-
{
|
| 412 |
-
"arm": "task_agnostic_base",
|
| 413 |
-
"label": "Black-and-White_Fallacy",
|
| 414 |
-
"support": 12,
|
| 415 |
-
"precision_mean": 0.3667929292929293,
|
| 416 |
-
"precision_sample_sd": 0.07670649392711647,
|
| 417 |
-
"recall_mean": 0.47222222222222227,
|
| 418 |
-
"recall_sample_sd": 0.09622504486493764,
|
| 419 |
-
"f1_mean": 0.4053892966936445,
|
| 420 |
-
"f1_sample_sd": 0.04211369218048746
|
| 421 |
-
},
|
| 422 |
-
{
|
| 423 |
-
"arm": "task_agnostic_base",
|
| 424 |
-
"label": "Causal_Oversimplification",
|
| 425 |
-
"support": 18,
|
| 426 |
-
"precision_mean": 0.5428824049513704,
|
| 427 |
-
"precision_sample_sd": 0.039216350192122156,
|
| 428 |
-
"recall_mean": 0.7962962962962963,
|
| 429 |
-
"recall_sample_sd": 0.11564811108145183,
|
| 430 |
-
"f1_mean": 0.6446992723588468,
|
| 431 |
-
"f1_sample_sd": 0.0634561308339175
|
| 432 |
-
},
|
| 433 |
-
{
|
| 434 |
-
"arm": "task_agnostic_base",
|
| 435 |
-
"label": "Doubt",
|
| 436 |
-
"support": 22,
|
| 437 |
-
"precision_mean": 0.5223832528180354,
|
| 438 |
-
"precision_sample_sd": 0.06742780050556177,
|
| 439 |
-
"recall_mean": 0.9848484848484849,
|
| 440 |
-
"recall_sample_sd": 0.026243194054073875,
|
| 441 |
-
"f1_mean": 0.6802057822343391,
|
| 442 |
-
"f1_sample_sd": 0.04928561754686331
|
| 443 |
-
},
|
| 444 |
-
{
|
| 445 |
-
"arm": "task_agnostic_base",
|
| 446 |
-
"label": "Exaggeration,Minimisation",
|
| 447 |
-
"support": 22,
|
| 448 |
-
"precision_mean": 0.4620811287477954,
|
| 449 |
-
"precision_sample_sd": 0.016801198309751,
|
| 450 |
-
"recall_mean": 0.7727272727272727,
|
| 451 |
-
"recall_sample_sd": 0.15745916432444337,
|
| 452 |
-
"f1_mean": 0.5727040816326531,
|
| 453 |
-
"f1_sample_sd": 0.03645259990419198
|
| 454 |
-
},
|
| 455 |
-
{
|
| 456 |
-
"arm": "task_agnostic_base",
|
| 457 |
-
"label": "Flag-Waving",
|
| 458 |
-
"support": 17,
|
| 459 |
-
"precision_mean": 0.5642642642642642,
|
| 460 |
-
"precision_sample_sd": 0.1478145677656334,
|
| 461 |
-
"recall_mean": 0.8235294117647058,
|
| 462 |
-
"recall_sample_sd": 0.1764705882352941,
|
| 463 |
-
"f1_mean": 0.6464506172839506,
|
| 464 |
-
"f1_sample_sd": 0.035742219952758225
|
| 465 |
-
},
|
| 466 |
-
{
|
| 467 |
-
"arm": "task_agnostic_base",
|
| 468 |
-
"label": "Loaded_Language",
|
| 469 |
-
"support": 45,
|
| 470 |
-
"precision_mean": 0.8649032907793016,
|
| 471 |
-
"precision_sample_sd": 0.046984312428427724,
|
| 472 |
-
"recall_mean": 0.9851851851851852,
|
| 473 |
-
"recall_sample_sd": 0.012830005981991702,
|
| 474 |
-
"f1_mean": 0.9208741623431562,
|
| 475 |
-
"f1_sample_sd": 0.03200926878737141
|
| 476 |
-
},
|
| 477 |
-
{
|
| 478 |
-
"arm": "task_agnostic_base",
|
| 479 |
-
"label": "Name_Calling,Labeling",
|
| 480 |
-
"support": 33,
|
| 481 |
-
"precision_mean": 0.7380952380952381,
|
| 482 |
-
"precision_sample_sd": 0.0047619047619048005,
|
| 483 |
-
"recall_mean": 0.9090909090909091,
|
| 484 |
-
"recall_sample_sd": 0.10925912955951485,
|
| 485 |
-
"f1_mean": 0.8125087983911513,
|
| 486 |
-
"f1_sample_sd": 0.04252971464057227
|
| 487 |
-
},
|
| 488 |
-
{
|
| 489 |
-
"arm": "task_agnostic_base",
|
| 490 |
-
"label": "Repetition",
|
| 491 |
-
"support": 23,
|
| 492 |
-
"precision_mean": 0.5196620583717357,
|
| 493 |
-
"precision_sample_sd": 0.021645909050194662,
|
| 494 |
-
"recall_mean": 0.7536231884057971,
|
| 495 |
-
"recall_sample_sd": 0.06641414050660639,
|
| 496 |
-
"f1_mean": 0.6147520781519571,
|
| 497 |
-
"f1_sample_sd": 0.03505927109416221
|
| 498 |
-
},
|
| 499 |
-
{
|
| 500 |
-
"arm": "task_agnostic_base",
|
| 501 |
-
"label": "Slogans",
|
| 502 |
-
"support": 14,
|
| 503 |
-
"precision_mean": 0.5161361626878869,
|
| 504 |
-
"precision_sample_sd": 0.20305677930281646,
|
| 505 |
-
"recall_mean": 0.7380952380952381,
|
| 506 |
-
"recall_sample_sd": 0.10910894511799614,
|
| 507 |
-
"f1_mean": 0.5834824090638044,
|
| 508 |
-
"f1_sample_sd": 0.09862686289220693
|
| 509 |
-
},
|
| 510 |
-
{
|
| 511 |
-
"arm": "task_agnostic_base",
|
| 512 |
-
"label": "Thought-terminating_Cliches",
|
| 513 |
-
"support": 7,
|
| 514 |
-
"precision_mean": 0.3492063492063492,
|
| 515 |
-
"precision_sample_sd": 0.143516994603374,
|
| 516 |
-
"recall_mean": 0.3333333333333333,
|
| 517 |
-
"recall_sample_sd": 0.08247860988423225,
|
| 518 |
-
"f1_mean": 0.319014319014319,
|
| 519 |
-
"f1_sample_sd": 0.04017590711117064
|
| 520 |
-
},
|
| 521 |
-
{
|
| 522 |
-
"arm": "task_agnostic_base",
|
| 523 |
-
"label": "Whataboutism,Straw_Men,Red_Herring",
|
| 524 |
-
"support": 10,
|
| 525 |
-
"precision_mean": 0.2761904761904762,
|
| 526 |
-
"precision_sample_sd": 0.05302632726504782,
|
| 527 |
-
"recall_mean": 0.5,
|
| 528 |
-
"recall_sample_sd": 0.2645751311064591,
|
| 529 |
-
"f1_mean": 0.330448343079922,
|
| 530 |
-
"f1_sample_sd": 0.021845164720973014
|
| 531 |
-
},
|
| 532 |
-
{
|
| 533 |
-
"arm": "structural_copy_control",
|
| 534 |
-
"label": "Appeal_to_Authority",
|
| 535 |
-
"support": 11,
|
| 536 |
-
"precision_mean": 0.26003086419753085,
|
| 537 |
-
"precision_sample_sd": 0.06281357095350094,
|
| 538 |
-
"recall_mean": 0.5757575757575758,
|
| 539 |
-
"recall_sample_sd": 0.262431940540739,
|
| 540 |
-
"f1_mean": 0.35477582846003897,
|
| 541 |
-
"f1_sample_sd": 0.11479479036518875
|
| 542 |
-
},
|
| 543 |
-
{
|
| 544 |
-
"arm": "structural_copy_control",
|
| 545 |
-
"label": "Appeal_to_fear-prejudice",
|
| 546 |
-
"support": 24,
|
| 547 |
-
"precision_mean": 0.5104813315339631,
|
| 548 |
-
"precision_sample_sd": 0.038427508996734555,
|
| 549 |
-
"recall_mean": 0.8611111111111112,
|
| 550 |
-
"recall_sample_sd": 0.024056261216234387,
|
| 551 |
-
"f1_mean": 0.6401745363877202,
|
| 552 |
-
"f1_sample_sd": 0.029305467347150463
|
| 553 |
-
},
|
| 554 |
-
{
|
| 555 |
-
"arm": "structural_copy_control",
|
| 556 |
-
"label": "Bandwagon,Reductio_ad_hitlerum",
|
| 557 |
-
"support": 7,
|
| 558 |
-
"precision_mean": 0.2876623376623376,
|
| 559 |
-
"precision_sample_sd": 0.061386823077618426,
|
| 560 |
-
"recall_mean": 0.8571428571428571,
|
| 561 |
-
"recall_sample_sd": 0.14285714285714285,
|
| 562 |
-
"f1_mean": 0.4306391777656145,
|
| 563 |
-
"f1_sample_sd": 0.08686392624022733
|
| 564 |
-
},
|
| 565 |
-
{
|
| 566 |
-
"arm": "structural_copy_control",
|
| 567 |
-
"label": "Black-and-White_Fallacy",
|
| 568 |
-
"support": 12,
|
| 569 |
-
"precision_mean": 0.31339031339031337,
|
| 570 |
-
"precision_sample_sd": 0.08301311842925053,
|
| 571 |
-
"recall_mean": 0.5277777777777778,
|
| 572 |
-
"recall_sample_sd": 0.2678791878053599,
|
| 573 |
-
"f1_mean": 0.38778427550357375,
|
| 574 |
-
"f1_sample_sd": 0.13069755090515112
|
| 575 |
-
},
|
| 576 |
-
{
|
| 577 |
-
"arm": "structural_copy_control",
|
| 578 |
-
"label": "Causal_Oversimplification",
|
| 579 |
-
"support": 18,
|
| 580 |
-
"precision_mean": 0.4464201043148412,
|
| 581 |
-
"precision_sample_sd": 0.013538446216471536,
|
| 582 |
-
"recall_mean": 0.9259259259259259,
|
| 583 |
-
"recall_sample_sd": 0.03207501495497923,
|
| 584 |
-
"f1_mean": 0.6023809523809524,
|
| 585 |
-
"f1_sample_sd": 0.018643640071557464
|
| 586 |
-
},
|
| 587 |
-
{
|
| 588 |
-
"arm": "structural_copy_control",
|
| 589 |
-
"label": "Doubt",
|
| 590 |
-
"support": 22,
|
| 591 |
-
"precision_mean": 0.4455314009661836,
|
| 592 |
-
"precision_sample_sd": 0.04067072694857813,
|
| 593 |
-
"recall_mean": 1.0,
|
| 594 |
-
"recall_sample_sd": 0.0,
|
| 595 |
-
"f1_mean": 0.615686274509804,
|
| 596 |
-
"f1_sample_sd": 0.039427200694482675
|
| 597 |
-
},
|
| 598 |
-
{
|
| 599 |
-
"arm": "structural_copy_control",
|
| 600 |
-
"label": "Exaggeration,Minimisation",
|
| 601 |
-
"support": 22,
|
| 602 |
-
"precision_mean": 0.4364176150420584,
|
| 603 |
-
"precision_sample_sd": 0.025093958106097894,
|
| 604 |
-
"recall_mean": 0.8787878787878788,
|
| 605 |
-
"recall_sample_sd": 0.05248638810814775,
|
| 606 |
-
"f1_mean": 0.5829803047194351,
|
| 607 |
-
"f1_sample_sd": 0.030899288356345502
|
| 608 |
-
},
|
| 609 |
-
{
|
| 610 |
-
"arm": "structural_copy_control",
|
| 611 |
-
"label": "Flag-Waving",
|
| 612 |
-
"support": 17,
|
| 613 |
-
"precision_mean": 0.3814814814814815,
|
| 614 |
-
"precision_sample_sd": 0.05010277503136551,
|
| 615 |
-
"recall_mean": 0.803921568627451,
|
| 616 |
-
"recall_sample_sd": 0.17970885078258195,
|
| 617 |
-
"f1_mean": 0.5138595287119652,
|
| 618 |
-
"f1_sample_sd": 0.06400931989326974
|
| 619 |
-
},
|
| 620 |
-
{
|
| 621 |
-
"arm": "structural_copy_control",
|
| 622 |
-
"label": "Loaded_Language",
|
| 623 |
-
"support": 45,
|
| 624 |
-
"precision_mean": 0.8492581841638446,
|
| 625 |
-
"precision_sample_sd": 0.016026591848331686,
|
| 626 |
-
"recall_mean": 1.0,
|
| 627 |
-
"recall_sample_sd": 0.0,
|
| 628 |
-
"f1_mean": 0.9184311025253588,
|
| 629 |
-
"f1_sample_sd": 0.009372233868132662
|
| 630 |
-
},
|
| 631 |
-
{
|
| 632 |
-
"arm": "structural_copy_control",
|
| 633 |
-
"label": "Name_Calling,Labeling",
|
| 634 |
-
"support": 33,
|
| 635 |
-
"precision_mean": 0.6375,
|
| 636 |
-
"precision_sample_sd": 0.03410563654946854,
|
| 637 |
-
"recall_mean": 0.9696969696969697,
|
| 638 |
-
"recall_sample_sd": 0.030303030303030276,
|
| 639 |
-
"f1_mean": 0.7685185185185185,
|
| 640 |
-
"f1_sample_sd": 0.02023900779105559
|
| 641 |
-
},
|
| 642 |
-
{
|
| 643 |
-
"arm": "structural_copy_control",
|
| 644 |
-
"label": "Repetition",
|
| 645 |
-
"support": 23,
|
| 646 |
-
"precision_mean": 0.44863933452168747,
|
| 647 |
-
"precision_sample_sd": 0.01520014564765896,
|
| 648 |
-
"recall_mean": 0.9420289855072463,
|
| 649 |
-
"recall_sample_sd": 0.06641414050660639,
|
| 650 |
-
"f1_mean": 0.607248835423033,
|
| 651 |
-
"f1_sample_sd": 0.019858627551810427
|
| 652 |
-
},
|
| 653 |
-
{
|
| 654 |
-
"arm": "structural_copy_control",
|
| 655 |
-
"label": "Slogans",
|
| 656 |
-
"support": 14,
|
| 657 |
-
"precision_mean": 0.3274318274318274,
|
| 658 |
-
"precision_sample_sd": 0.05434598572773684,
|
| 659 |
-
"recall_mean": 0.4523809523809524,
|
| 660 |
-
"recall_sample_sd": 0.1486904285332952,
|
| 661 |
-
"f1_mean": 0.3715828924162258,
|
| 662 |
-
"f1_sample_sd": 0.07106660236343577
|
| 663 |
-
},
|
| 664 |
-
{
|
| 665 |
-
"arm": "structural_copy_control",
|
| 666 |
-
"label": "Thought-terminating_Cliches",
|
| 667 |
-
"support": 7,
|
| 668 |
-
"precision_mean": 0.1142156862745098,
|
| 669 |
-
"precision_sample_sd": 0.012848374923608973,
|
| 670 |
-
"recall_mean": 0.38095238095238093,
|
| 671 |
-
"recall_sample_sd": 0.1649572197684645,
|
| 672 |
-
"f1_mean": 0.17331433998100665,
|
| 673 |
-
"f1_sample_sd": 0.029065880233243847
|
| 674 |
-
},
|
| 675 |
-
{
|
| 676 |
-
"arm": "structural_copy_control",
|
| 677 |
-
"label": "Whataboutism,Straw_Men,Red_Herring",
|
| 678 |
-
"support": 10,
|
| 679 |
-
"precision_mean": 0.19344598054275472,
|
| 680 |
-
"precision_sample_sd": 0.03373872471827226,
|
| 681 |
-
"recall_mean": 0.5666666666666667,
|
| 682 |
-
"recall_sample_sd": 0.2081665999466133,
|
| 683 |
-
"f1_mean": 0.28584083690299267,
|
| 684 |
-
"f1_sample_sd": 0.06078854830133676
|
| 685 |
-
},
|
| 686 |
-
{
|
| 687 |
-
"arm": "existing_specialized_separate_lineage",
|
| 688 |
-
"label": "Appeal_to_Authority",
|
| 689 |
-
"support": 11,
|
| 690 |
-
"precision_mean": 0.37836021505376344,
|
| 691 |
-
"precision_sample_sd": 0.10546365636105155,
|
| 692 |
-
"recall_mean": 0.6666666666666666,
|
| 693 |
-
"recall_sample_sd": 0.2287828616748712,
|
| 694 |
-
"f1_mean": 0.46885361552028215,
|
| 695 |
-
"f1_sample_sd": 0.09502747629890008
|
| 696 |
-
},
|
| 697 |
-
{
|
| 698 |
-
"arm": "existing_specialized_separate_lineage",
|
| 699 |
-
"label": "Appeal_to_fear-prejudice",
|
| 700 |
-
"support": 24,
|
| 701 |
-
"precision_mean": 0.525708635464733,
|
| 702 |
-
"precision_sample_sd": 0.02235198659615819,
|
| 703 |
-
"recall_mean": 0.8888888888888888,
|
| 704 |
-
"recall_sample_sd": 0.048112522432468774,
|
| 705 |
-
"f1_mean": 0.6599065351235072,
|
| 706 |
-
"f1_sample_sd": 0.015362381542320084
|
| 707 |
-
},
|
| 708 |
-
{
|
| 709 |
-
"arm": "existing_specialized_separate_lineage",
|
| 710 |
-
"label": "Bandwagon,Reductio_ad_hitlerum",
|
| 711 |
-
"support": 7,
|
| 712 |
-
"precision_mean": 0.44485144485144484,
|
| 713 |
-
"precision_sample_sd": 0.016487284575115283,
|
| 714 |
-
"recall_mean": 0.7619047619047619,
|
| 715 |
-
"recall_sample_sd": 0.1649572197684645,
|
| 716 |
-
"f1_mean": 0.5571428571428572,
|
| 717 |
-
"f1_sample_sd": 0.051507875363771265
|
| 718 |
-
},
|
| 719 |
-
{
|
| 720 |
-
"arm": "existing_specialized_separate_lineage",
|
| 721 |
-
"label": "Black-and-White_Fallacy",
|
| 722 |
-
"support": 12,
|
| 723 |
-
"precision_mean": 0.4662644544997486,
|
| 724 |
-
"precision_sample_sd": 0.004538274146363688,
|
| 725 |
-
"recall_mean": 0.5833333333333334,
|
| 726 |
-
"recall_sample_sd": 0.08333333333333331,
|
| 727 |
-
"f1_mean": 0.516747552149851,
|
| 728 |
-
"f1_sample_sd": 0.03589484965697345
|
| 729 |
-
},
|
| 730 |
-
{
|
| 731 |
-
"arm": "existing_specialized_separate_lineage",
|
| 732 |
-
"label": "Causal_Oversimplification",
|
| 733 |
-
"support": 18,
|
| 734 |
-
"precision_mean": 0.4472934472934473,
|
| 735 |
-
"precision_sample_sd": 0.013055771210700413,
|
| 736 |
-
"recall_mean": 0.8333333333333333,
|
| 737 |
-
"recall_sample_sd": 0.1469861839480328,
|
| 738 |
-
"f1_mean": 0.5781794553724379,
|
| 739 |
-
"f1_sample_sd": 0.028407563015813122
|
| 740 |
-
},
|
| 741 |
-
{
|
| 742 |
-
"arm": "existing_specialized_separate_lineage",
|
| 743 |
-
"label": "Doubt",
|
| 744 |
-
"support": 22,
|
| 745 |
-
"precision_mean": 0.5045661001171619,
|
| 746 |
-
"precision_sample_sd": 0.023581174969843577,
|
| 747 |
-
"recall_mean": 1.0,
|
| 748 |
-
"recall_sample_sd": 0.0,
|
| 749 |
-
"f1_mean": 0.6704939668174963,
|
| 750 |
-
"f1_sample_sd": 0.020973132363659372
|
| 751 |
-
},
|
| 752 |
-
{
|
| 753 |
-
"arm": "existing_specialized_separate_lineage",
|
| 754 |
-
"label": "Exaggeration,Minimisation",
|
| 755 |
-
"support": 22,
|
| 756 |
-
"precision_mean": 0.44589793281653745,
|
| 757 |
-
"precision_sample_sd": 0.010987847993006481,
|
| 758 |
-
"recall_mean": 0.9393939393939394,
|
| 759 |
-
"recall_sample_sd": 0.06943296507508849,
|
| 760 |
-
"f1_mean": 0.6043956043956044,
|
| 761 |
-
"f1_sample_sd": 0.022305256187015846
|
| 762 |
-
},
|
| 763 |
-
{
|
| 764 |
-
"arm": "existing_specialized_separate_lineage",
|
| 765 |
-
"label": "Flag-Waving",
|
| 766 |
-
"support": 17,
|
| 767 |
-
"precision_mean": 0.5576388888888889,
|
| 768 |
-
"precision_sample_sd": 0.03705367103320583,
|
| 769 |
-
"recall_mean": 0.8823529411764706,
|
| 770 |
-
"recall_sample_sd": 0.11764705882352944,
|
| 771 |
-
"f1_mean": 0.680769868923179,
|
| 772 |
-
"f1_sample_sd": 0.04164658769406446
|
| 773 |
-
},
|
| 774 |
-
{
|
| 775 |
-
"arm": "existing_specialized_separate_lineage",
|
| 776 |
-
"label": "Loaded_Language",
|
| 777 |
-
"support": 45,
|
| 778 |
-
"precision_mean": 0.8387650085763294,
|
| 779 |
-
"precision_sample_sd": 0.01782556577257849,
|
| 780 |
-
"recall_mean": 1.0,
|
| 781 |
-
"recall_sample_sd": 0.0,
|
| 782 |
-
"f1_mean": 0.9122448979591837,
|
| 783 |
-
"f1_sample_sd": 0.01060439269940129
|
| 784 |
-
},
|
| 785 |
-
{
|
| 786 |
-
"arm": "existing_specialized_separate_lineage",
|
| 787 |
-
"label": "Name_Calling,Labeling",
|
| 788 |
-
"support": 33,
|
| 789 |
-
"precision_mean": 0.6957104190221673,
|
| 790 |
-
"precision_sample_sd": 0.019201490809269257,
|
| 791 |
-
"recall_mean": 0.9696969696969697,
|
| 792 |
-
"recall_sample_sd": 0.030303030303030276,
|
| 793 |
-
"f1_mean": 0.8101049442821594,
|
| 794 |
-
"f1_sample_sd": 0.022092733328802537
|
| 795 |
-
},
|
| 796 |
-
{
|
| 797 |
-
"arm": "existing_specialized_separate_lineage",
|
| 798 |
-
"label": "Repetition",
|
| 799 |
-
"support": 23,
|
| 800 |
-
"precision_mean": 0.4781557067271353,
|
| 801 |
-
"precision_sample_sd": 0.009897987322673213,
|
| 802 |
-
"recall_mean": 0.9420289855072463,
|
| 803 |
-
"recall_sample_sd": 0.06641414050660639,
|
| 804 |
-
"f1_mean": 0.6337774426009719,
|
| 805 |
-
"f1_sample_sd": 0.016444118333485254
|
| 806 |
-
},
|
| 807 |
-
{
|
| 808 |
-
"arm": "existing_specialized_separate_lineage",
|
| 809 |
-
"label": "Slogans",
|
| 810 |
-
"support": 14,
|
| 811 |
-
"precision_mean": 0.4334215167548501,
|
| 812 |
-
"precision_sample_sd": 0.05811581861282867,
|
| 813 |
-
"recall_mean": 0.6904761904761905,
|
| 814 |
-
"recall_sample_sd": 0.1649572197684645,
|
| 815 |
-
"f1_mean": 0.5201316298877274,
|
| 816 |
-
"f1_sample_sd": 0.018567913895656114
|
| 817 |
-
},
|
| 818 |
-
{
|
| 819 |
-
"arm": "existing_specialized_separate_lineage",
|
| 820 |
-
"label": "Thought-terminating_Cliches",
|
| 821 |
-
"support": 7,
|
| 822 |
-
"precision_mean": 0.18446115288220552,
|
| 823 |
-
"precision_sample_sd": 0.03641251252864963,
|
| 824 |
-
"recall_mean": 0.6666666666666666,
|
| 825 |
-
"recall_sample_sd": 0.1649572197684645,
|
| 826 |
-
"f1_mean": 0.2868626868626869,
|
| 827 |
-
"f1_sample_sd": 0.0511620864429468
|
| 828 |
-
},
|
| 829 |
-
{
|
| 830 |
-
"arm": "existing_specialized_separate_lineage",
|
| 831 |
-
"label": "Whataboutism,Straw_Men,Red_Herring",
|
| 832 |
-
"support": 10,
|
| 833 |
-
"precision_mean": 0.2777777777777778,
|
| 834 |
-
"precision_sample_sd": 0.0481125224324688,
|
| 835 |
-
"recall_mean": 0.4,
|
| 836 |
-
"recall_sample_sd": 0.0,
|
| 837 |
-
"f1_mean": 0.32634032634032634,
|
| 838 |
-
"f1_sample_sd": 0.03229931575886014
|
| 839 |
-
}
|
| 840 |
-
],
|
| 841 |
-
"compression_ladder": [
|
| 842 |
-
{
|
| 843 |
-
"name": "24L upstream text teacher",
|
| 844 |
-
"layers": 24,
|
| 845 |
-
"parameters": 752393024,
|
| 846 |
-
"weight_bytes": 1504825632,
|
| 847 |
-
"weight_mib": 1435.1135559082031,
|
| 848 |
-
"weight_sha256": "30783da4651259c26b8a204821ea426b9526431e0c14da0d1c6fa284e7414e5e",
|
| 849 |
-
"kd_stage": "",
|
| 850 |
-
"mean_total_loss": "",
|
| 851 |
-
"mean_interface_loss": "",
|
| 852 |
-
"mean_final_loss": ""
|
| 853 |
-
},
|
| 854 |
-
{
|
| 855 |
-
"name": "8L task-free KD",
|
| 856 |
-
"layers": 8,
|
| 857 |
-
"parameters": 420318144,
|
| 858 |
-
"weight_bytes": 840647880,
|
| 859 |
-
"weight_mib": 801.7042922973633,
|
| 860 |
-
"weight_sha256": "086fbaf9a4838ecde56cd3541c9c3212b22d0c5ee745cf2c1c5e0bb31e0ec22c",
|
| 861 |
-
"kd_stage": "24to8",
|
| 862 |
-
"mean_total_loss": 0.17622653172702485,
|
| 863 |
-
"mean_interface_loss": 0.21562811763578793,
|
| 864 |
-
"mean_final_loss": 0.16626427527853593
|
| 865 |
-
},
|
| 866 |
-
{
|
| 867 |
-
"name": "6L task-free KD",
|
| 868 |
-
"layers": 6,
|
| 869 |
-
"parameters": 377207424,
|
| 870 |
-
"weight_bytes": 754423424,
|
| 871 |
-
"weight_mib": 719.4742431640625,
|
| 872 |
-
"weight_sha256": "492749c20f77a42f6bc993f0927a82217b94f872b7eacf25612f7520c9bc28ce",
|
| 873 |
-
"kd_stage": "8to6",
|
| 874 |
-
"mean_total_loss": 0.03382104352249371,
|
| 875 |
-
"mean_interface_loss": 0.04486069045515251,
|
| 876 |
-
"mean_final_loss": 0.031147060785087888
|
| 877 |
-
},
|
| 878 |
-
{
|
| 879 |
-
"name": "4L task-free KD root",
|
| 880 |
-
"layers": 4,
|
| 881 |
-
"parameters": 334096704,
|
| 882 |
-
"weight_bytes": 668198976,
|
| 883 |
-
"weight_mib": 637.2442016601562,
|
| 884 |
-
"weight_sha256": "2732c616772fe320cdea228ab4554981418b1b2bf615c4183fb1ac8e6e2168d3",
|
| 885 |
-
"kd_stage": "6to4",
|
| 886 |
-
"mean_total_loss": 0.047020394468859195,
|
| 887 |
-
"mean_interface_loss": 0.07048421185527332,
|
| 888 |
-
"mean_final_loss": 0.043119939065377366
|
| 889 |
-
}
|
| 890 |
-
]
|
| 891 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@@ -1,891 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"schema": "standalone4l-complete-transfer-benchmark-v2",
|
| 3 |
-
"status": "complete_from_frozen_reports",
|
| 4 |
-
"claim_scope": "Exploratory fresh-head transfer evidence on one English task and a previously opened 55-article SemEval-derived test split.",
|
| 5 |
-
"headless_root_requires_adaptation": true,
|
| 6 |
-
"test_previously_opened": true,
|
| 7 |
-
"test_articles": 55,
|
| 8 |
-
"test_windows": 434,
|
| 9 |
-
"seeds": [
|
| 10 |
-
41,
|
| 11 |
-
42,
|
| 12 |
-
43
|
| 13 |
-
],
|
| 14 |
-
"three_seed_sd_is_not_confidence_interval": true,
|
| 15 |
-
"multilingual_downstream_evaluation_performed": false,
|
| 16 |
-
"upstream_multilingual_metadata": {
|
| 17 |
-
"source": "Qwen/Qwen3.5-0.8B official model card",
|
| 18 |
-
"languages_and_dialects_claimed_upstream": 201,
|
| 19 |
-
"scope": "inherited tokenizer and architecture coverage only"
|
| 20 |
-
},
|
| 21 |
-
"protocol": {
|
| 22 |
-
"epochs": 5,
|
| 23 |
-
"primary_epoch": 5,
|
| 24 |
-
"early_stopping": false,
|
| 25 |
-
"max_length": 256,
|
| 26 |
-
"stride": 128,
|
| 27 |
-
"window_target": "span-aligned technique labels",
|
| 28 |
-
"article_aggregation": "labelwise maximum probability",
|
| 29 |
-
"micro_batch_size": 1,
|
| 30 |
-
"gradient_accumulation": 32,
|
| 31 |
-
"effective_batch_size": 32,
|
| 32 |
-
"learning_rate": 2e-05,
|
| 33 |
-
"weight_decay": 0.01,
|
| 34 |
-
"gradient_clip": 1.0,
|
| 35 |
-
"optimizer": "AdamW",
|
| 36 |
-
"dtype": "bfloat16",
|
| 37 |
-
"thresholds": [
|
| 38 |
-
0.05,
|
| 39 |
-
0.1,
|
| 40 |
-
0.15,
|
| 41 |
-
0.2,
|
| 42 |
-
0.25,
|
| 43 |
-
0.3,
|
| 44 |
-
0.35,
|
| 45 |
-
0.4,
|
| 46 |
-
0.45,
|
| 47 |
-
0.5,
|
| 48 |
-
0.55,
|
| 49 |
-
0.6,
|
| 50 |
-
0.65,
|
| 51 |
-
0.7,
|
| 52 |
-
0.75,
|
| 53 |
-
0.8,
|
| 54 |
-
0.85,
|
| 55 |
-
0.9
|
| 56 |
-
],
|
| 57 |
-
"threshold_selection": "epoch-5 calibration macro-F1, then micro-F1"
|
| 58 |
-
},
|
| 59 |
-
"arms": {
|
| 60 |
-
"task_agnostic_base": {
|
| 61 |
-
"display": "Task-agnostic KD base",
|
| 62 |
-
"lineage": "root base plus a fresh 14-label transfer head",
|
| 63 |
-
"metrics": {
|
| 64 |
-
"macro_f1": {
|
| 65 |
-
"values": [
|
| 66 |
-
0.5883093293517536,
|
| 67 |
-
0.6028379036218096,
|
| 68 |
-
0.5712816197954101
|
| 69 |
-
],
|
| 70 |
-
"mean": 0.5874762842563245,
|
| 71 |
-
"sample_sd": 0.01579462678670608,
|
| 72 |
-
"n": 3
|
| 73 |
-
},
|
| 74 |
-
"micro_f1": {
|
| 75 |
-
"values": [
|
| 76 |
-
0.6468085106382978,
|
| 77 |
-
0.6655172413793103,
|
| 78 |
-
0.6465256797583082
|
| 79 |
-
],
|
| 80 |
-
"mean": 0.6529504772586388,
|
| 81 |
-
"sample_sd": 0.010884055708931178,
|
| 82 |
-
"n": 3
|
| 83 |
-
},
|
| 84 |
-
"exact_match": {
|
| 85 |
-
"values": [
|
| 86 |
-
0.05454545454545454,
|
| 87 |
-
0.03636363636363636,
|
| 88 |
-
0.0
|
| 89 |
-
],
|
| 90 |
-
"mean": 0.0303030303030303,
|
| 91 |
-
"sample_sd": 0.027773186030035393,
|
| 92 |
-
"n": 3
|
| 93 |
-
},
|
| 94 |
-
"predicted_positive_rate": {
|
| 95 |
-
"values": [
|
| 96 |
-
0.5714285714285714,
|
| 97 |
-
0.4090909090909091,
|
| 98 |
-
0.5155844155844156
|
| 99 |
-
],
|
| 100 |
-
"mean": 0.4987012987012987,
|
| 101 |
-
"sample_sd": 0.08247520160679181,
|
| 102 |
-
"n": 3
|
| 103 |
-
}
|
| 104 |
-
},
|
| 105 |
-
"resources": {
|
| 106 |
-
"parameters": {
|
| 107 |
-
"values": [
|
| 108 |
-
334111040.0,
|
| 109 |
-
334111040.0,
|
| 110 |
-
334111040.0
|
| 111 |
-
],
|
| 112 |
-
"mean": 334111040.0,
|
| 113 |
-
"sample_sd": 0.0,
|
| 114 |
-
"n": 3
|
| 115 |
-
},
|
| 116 |
-
"peak_cuda_allocated_bytes": {
|
| 117 |
-
"values": [
|
| 118 |
-
3380975616.0,
|
| 119 |
-
3380975616.0,
|
| 120 |
-
3380975616.0
|
| 121 |
-
],
|
| 122 |
-
"mean": 3380975616.0,
|
| 123 |
-
"sample_sd": 0.0,
|
| 124 |
-
"n": 3
|
| 125 |
-
},
|
| 126 |
-
"peak_cuda_reserved_bytes": {
|
| 127 |
-
"values": [
|
| 128 |
-
3519021056.0,
|
| 129 |
-
3519021056.0,
|
| 130 |
-
3510632448.0
|
| 131 |
-
],
|
| 132 |
-
"mean": 3516224853.3333335,
|
| 133 |
-
"sample_sd": 4843165.086926248,
|
| 134 |
-
"n": 3
|
| 135 |
-
},
|
| 136 |
-
"total_wall_seconds": {
|
| 137 |
-
"values": [
|
| 138 |
-
1583.0297977919981,
|
| 139 |
-
1588.617994428001,
|
| 140 |
-
1603.4310500120046
|
| 141 |
-
],
|
| 142 |
-
"mean": 1591.692947410668,
|
| 143 |
-
"sample_sd": 10.54249851412311,
|
| 144 |
-
"n": 3
|
| 145 |
-
},
|
| 146 |
-
"test_seconds_per_article": {
|
| 147 |
-
"values": [
|
| 148 |
-
0.08164824154545468,
|
| 149 |
-
0.07871871667289683,
|
| 150 |
-
0.08355205721825107
|
| 151 |
-
],
|
| 152 |
-
"mean": 0.08130633847886752,
|
| 153 |
-
"sample_sd": 0.0024347419754560395,
|
| 154 |
-
"n": 3
|
| 155 |
-
}
|
| 156 |
-
},
|
| 157 |
-
"thresholds": [
|
| 158 |
-
0.15,
|
| 159 |
-
0.3,
|
| 160 |
-
0.1
|
| 161 |
-
]
|
| 162 |
-
},
|
| 163 |
-
"structural_copy_control": {
|
| 164 |
-
"display": "Structural copy, no task-free KD",
|
| 165 |
-
"lineage": "same four-layer structure without task-free KD",
|
| 166 |
-
"metrics": {
|
| 167 |
-
"macro_f1": {
|
| 168 |
-
"values": [
|
| 169 |
-
0.5169424579262757,
|
| 170 |
-
0.5349405779540175,
|
| 171 |
-
0.5023778364496152
|
| 172 |
-
],
|
| 173 |
-
"mean": 0.5180869574433028,
|
| 174 |
-
"sample_sd": 0.016311512588628545,
|
| 175 |
-
"n": 3
|
| 176 |
-
},
|
| 177 |
-
"micro_f1": {
|
| 178 |
-
"values": [
|
| 179 |
-
0.5773955773955775,
|
| 180 |
-
0.6083445491251682,
|
| 181 |
-
0.585635359116022
|
| 182 |
-
],
|
| 183 |
-
"mean": 0.5904584952122559,
|
| 184 |
-
"sample_sd": 0.01602830914762526,
|
| 185 |
-
"n": 3
|
| 186 |
-
},
|
| 187 |
-
"exact_match": {
|
| 188 |
-
"values": [
|
| 189 |
-
0.0,
|
| 190 |
-
0.0,
|
| 191 |
-
0.0
|
| 192 |
-
],
|
| 193 |
-
"mean": 0.0,
|
| 194 |
-
"sample_sd": 0.0,
|
| 195 |
-
"n": 3
|
| 196 |
-
},
|
| 197 |
-
"predicted_positive_rate": {
|
| 198 |
-
"values": [
|
| 199 |
-
0.712987012987013,
|
| 200 |
-
0.6207792207792208,
|
| 201 |
-
0.5961038961038961
|
| 202 |
-
],
|
| 203 |
-
"mean": 0.6432900432900432,
|
| 204 |
-
"sample_sd": 0.06160737456395461,
|
| 205 |
-
"n": 3
|
| 206 |
-
}
|
| 207 |
-
},
|
| 208 |
-
"resources": {
|
| 209 |
-
"parameters": {
|
| 210 |
-
"values": [
|
| 211 |
-
334111040.0,
|
| 212 |
-
334111040.0,
|
| 213 |
-
334111040.0
|
| 214 |
-
],
|
| 215 |
-
"mean": 334111040.0,
|
| 216 |
-
"sample_sd": 0.0,
|
| 217 |
-
"n": 3
|
| 218 |
-
},
|
| 219 |
-
"peak_cuda_allocated_bytes": {
|
| 220 |
-
"values": [
|
| 221 |
-
3380975616.0,
|
| 222 |
-
3380975616.0,
|
| 223 |
-
3380975616.0
|
| 224 |
-
],
|
| 225 |
-
"mean": 3380975616.0,
|
| 226 |
-
"sample_sd": 0.0,
|
| 227 |
-
"n": 3
|
| 228 |
-
},
|
| 229 |
-
"peak_cuda_reserved_bytes": {
|
| 230 |
-
"values": [
|
| 231 |
-
3519021056.0,
|
| 232 |
-
3519021056.0,
|
| 233 |
-
3510632448.0
|
| 234 |
-
],
|
| 235 |
-
"mean": 3516224853.3333335,
|
| 236 |
-
"sample_sd": 4843165.086926248,
|
| 237 |
-
"n": 3
|
| 238 |
-
},
|
| 239 |
-
"total_wall_seconds": {
|
| 240 |
-
"values": [
|
| 241 |
-
1575.5905925639963,
|
| 242 |
-
1623.5639530190092,
|
| 243 |
-
1629.195114912989
|
| 244 |
-
],
|
| 245 |
-
"mean": 1609.4498868319981,
|
| 246 |
-
"sample_sd": 29.457874198835384,
|
| 247 |
-
"n": 3
|
| 248 |
-
},
|
| 249 |
-
"test_seconds_per_article": {
|
| 250 |
-
"values": [
|
| 251 |
-
0.07877608058201573,
|
| 252 |
-
0.08307423260041767,
|
| 253 |
-
0.08163602005436339
|
| 254 |
-
],
|
| 255 |
-
"mean": 0.08116211107893226,
|
| 256 |
-
"sample_sd": 0.002187914527825257,
|
| 257 |
-
"n": 3
|
| 258 |
-
}
|
| 259 |
-
},
|
| 260 |
-
"thresholds": [
|
| 261 |
-
0.05,
|
| 262 |
-
0.2,
|
| 263 |
-
0.2
|
| 264 |
-
]
|
| 265 |
-
},
|
| 266 |
-
"existing_specialized_separate_lineage": {
|
| 267 |
-
"display": "Existing specialized 4L, separate lineage",
|
| 268 |
-
"lineage": "historical task-specific shrink/distillation lineage",
|
| 269 |
-
"metrics": {
|
| 270 |
-
"macro_f1": {
|
| 271 |
-
"values": [
|
| 272 |
-
0.5832848452288824,
|
| 273 |
-
0.5958405181160762,
|
| 274 |
-
0.5835785045218137
|
| 275 |
-
],
|
| 276 |
-
"mean": 0.5875679559555907,
|
| 277 |
-
"sample_sd": 0.007165753447433348,
|
| 278 |
-
"n": 3
|
| 279 |
-
},
|
| 280 |
-
"micro_f1": {
|
| 281 |
-
"values": [
|
| 282 |
-
0.6494252873563218,
|
| 283 |
-
0.6458616010854816,
|
| 284 |
-
0.6486486486486486
|
| 285 |
-
],
|
| 286 |
-
"mean": 0.647978512363484,
|
| 287 |
-
"sample_sd": 0.0018739735696546695,
|
| 288 |
-
"n": 3
|
| 289 |
-
},
|
| 290 |
-
"exact_match": {
|
| 291 |
-
"values": [
|
| 292 |
-
0.0,
|
| 293 |
-
0.0,
|
| 294 |
-
0.01818181818181818
|
| 295 |
-
],
|
| 296 |
-
"mean": 0.006060606060606061,
|
| 297 |
-
"sample_sd": 0.01049727762162956,
|
| 298 |
-
"n": 3
|
| 299 |
-
},
|
| 300 |
-
"predicted_positive_rate": {
|
| 301 |
-
"values": [
|
| 302 |
-
0.5597402597402598,
|
| 303 |
-
0.612987012987013,
|
| 304 |
-
0.5688311688311688
|
| 305 |
-
],
|
| 306 |
-
"mean": 0.5805194805194805,
|
| 307 |
-
"sample_sd": 0.02848274311618351,
|
| 308 |
-
"n": 3
|
| 309 |
-
}
|
| 310 |
-
},
|
| 311 |
-
"resources": {
|
| 312 |
-
"parameters": {
|
| 313 |
-
"values": [
|
| 314 |
-
210903360.0,
|
| 315 |
-
210903360.0,
|
| 316 |
-
210903360.0
|
| 317 |
-
],
|
| 318 |
-
"mean": 210903360.0,
|
| 319 |
-
"sample_sd": 0.0,
|
| 320 |
-
"n": 3
|
| 321 |
-
},
|
| 322 |
-
"peak_cuda_allocated_bytes": {
|
| 323 |
-
"values": [
|
| 324 |
-
2143655936.0,
|
| 325 |
-
2143655936.0,
|
| 326 |
-
2143655936.0
|
| 327 |
-
],
|
| 328 |
-
"mean": 2143655936.0,
|
| 329 |
-
"sample_sd": 0.0,
|
| 330 |
-
"n": 3
|
| 331 |
-
},
|
| 332 |
-
"peak_cuda_reserved_bytes": {
|
| 333 |
-
"values": [
|
| 334 |
-
2281701376.0,
|
| 335 |
-
2281701376.0,
|
| 336 |
-
2273312768.0
|
| 337 |
-
],
|
| 338 |
-
"mean": 2278905173.3333335,
|
| 339 |
-
"sample_sd": 4843165.086926248,
|
| 340 |
-
"n": 3
|
| 341 |
-
},
|
| 342 |
-
"total_wall_seconds": {
|
| 343 |
-
"values": [
|
| 344 |
-
1634.488557530014,
|
| 345 |
-
1652.9397718419787,
|
| 346 |
-
1643.488202479988
|
| 347 |
-
],
|
| 348 |
-
"mean": 1643.6388439506602,
|
| 349 |
-
"sample_sd": 9.226529522850198,
|
| 350 |
-
"n": 3
|
| 351 |
-
},
|
| 352 |
-
"test_seconds_per_article": {
|
| 353 |
-
"values": [
|
| 354 |
-
0.0847959352360869,
|
| 355 |
-
0.08598086021798239,
|
| 356 |
-
0.08552987641781908
|
| 357 |
-
],
|
| 358 |
-
"mean": 0.08543555729062946,
|
| 359 |
-
"sample_sd": 0.0005980667826383295,
|
| 360 |
-
"n": 3
|
| 361 |
-
}
|
| 362 |
-
},
|
| 363 |
-
"thresholds": [
|
| 364 |
-
0.1,
|
| 365 |
-
0.05,
|
| 366 |
-
0.05
|
| 367 |
-
]
|
| 368 |
-
}
|
| 369 |
-
},
|
| 370 |
-
"same_seed_macro_f1_deltas_base_minus_control": [
|
| 371 |
-
0.07136687142547793,
|
| 372 |
-
0.06789732566779205,
|
| 373 |
-
0.06890378334579494
|
| 374 |
-
],
|
| 375 |
-
"mean_macro_f1_delta_base_minus_control": 0.06938932681302164,
|
| 376 |
-
"publication_gate_passed": true,
|
| 377 |
-
"per_label_summary": [
|
| 378 |
-
{
|
| 379 |
-
"arm": "task_agnostic_base",
|
| 380 |
-
"label": "Appeal_to_Authority",
|
| 381 |
-
"support": 11,
|
| 382 |
-
"precision_mean": 0.3958614484930274,
|
| 383 |
-
"precision_sample_sd": 0.1238064457010119,
|
| 384 |
-
"recall_mean": 0.5757575757575757,
|
| 385 |
-
"recall_sample_sd": 0.05248638810814781,
|
| 386 |
-
"f1_mean": 0.4657088122605364,
|
| 387 |
-
"f1_sample_sd": 0.1020990126979069
|
| 388 |
-
},
|
| 389 |
-
{
|
| 390 |
-
"arm": "task_agnostic_base",
|
| 391 |
-
"label": "Appeal_to_fear-prejudice",
|
| 392 |
-
"support": 24,
|
| 393 |
-
"precision_mean": 0.5540244550770866,
|
| 394 |
-
"precision_sample_sd": 0.02525673990402998,
|
| 395 |
-
"recall_mean": 0.7361111111111112,
|
| 396 |
-
"recall_sample_sd": 0.13393959390267993,
|
| 397 |
-
"f1_mean": 0.6277521761392729,
|
| 398 |
-
"f1_sample_sd": 0.04994879016183189
|
| 399 |
-
},
|
| 400 |
-
{
|
| 401 |
-
"arm": "task_agnostic_base",
|
| 402 |
-
"label": "Bandwagon,Reductio_ad_hitlerum",
|
| 403 |
-
"support": 7,
|
| 404 |
-
"precision_mean": 0.4851851851851852,
|
| 405 |
-
"precision_sample_sd": 0.07882887649552875,
|
| 406 |
-
"recall_mean": 0.8095238095238095,
|
| 407 |
-
"recall_sample_sd": 0.08247860988423221,
|
| 408 |
-
"f1_mean": 0.6006778309409889,
|
| 409 |
-
"f1_sample_sd": 0.04793776258918466
|
| 410 |
-
},
|
| 411 |
-
{
|
| 412 |
-
"arm": "task_agnostic_base",
|
| 413 |
-
"label": "Black-and-White_Fallacy",
|
| 414 |
-
"support": 12,
|
| 415 |
-
"precision_mean": 0.3667929292929293,
|
| 416 |
-
"precision_sample_sd": 0.07670649392711647,
|
| 417 |
-
"recall_mean": 0.47222222222222227,
|
| 418 |
-
"recall_sample_sd": 0.09622504486493764,
|
| 419 |
-
"f1_mean": 0.4053892966936445,
|
| 420 |
-
"f1_sample_sd": 0.04211369218048746
|
| 421 |
-
},
|
| 422 |
-
{
|
| 423 |
-
"arm": "task_agnostic_base",
|
| 424 |
-
"label": "Causal_Oversimplification",
|
| 425 |
-
"support": 18,
|
| 426 |
-
"precision_mean": 0.5428824049513704,
|
| 427 |
-
"precision_sample_sd": 0.039216350192122156,
|
| 428 |
-
"recall_mean": 0.7962962962962963,
|
| 429 |
-
"recall_sample_sd": 0.11564811108145183,
|
| 430 |
-
"f1_mean": 0.6446992723588468,
|
| 431 |
-
"f1_sample_sd": 0.0634561308339175
|
| 432 |
-
},
|
| 433 |
-
{
|
| 434 |
-
"arm": "task_agnostic_base",
|
| 435 |
-
"label": "Doubt",
|
| 436 |
-
"support": 22,
|
| 437 |
-
"precision_mean": 0.5223832528180354,
|
| 438 |
-
"precision_sample_sd": 0.06742780050556177,
|
| 439 |
-
"recall_mean": 0.9848484848484849,
|
| 440 |
-
"recall_sample_sd": 0.026243194054073875,
|
| 441 |
-
"f1_mean": 0.6802057822343391,
|
| 442 |
-
"f1_sample_sd": 0.04928561754686331
|
| 443 |
-
},
|
| 444 |
-
{
|
| 445 |
-
"arm": "task_agnostic_base",
|
| 446 |
-
"label": "Exaggeration,Minimisation",
|
| 447 |
-
"support": 22,
|
| 448 |
-
"precision_mean": 0.4620811287477954,
|
| 449 |
-
"precision_sample_sd": 0.016801198309751,
|
| 450 |
-
"recall_mean": 0.7727272727272727,
|
| 451 |
-
"recall_sample_sd": 0.15745916432444337,
|
| 452 |
-
"f1_mean": 0.5727040816326531,
|
| 453 |
-
"f1_sample_sd": 0.03645259990419198
|
| 454 |
-
},
|
| 455 |
-
{
|
| 456 |
-
"arm": "task_agnostic_base",
|
| 457 |
-
"label": "Flag-Waving",
|
| 458 |
-
"support": 17,
|
| 459 |
-
"precision_mean": 0.5642642642642642,
|
| 460 |
-
"precision_sample_sd": 0.1478145677656334,
|
| 461 |
-
"recall_mean": 0.8235294117647058,
|
| 462 |
-
"recall_sample_sd": 0.1764705882352941,
|
| 463 |
-
"f1_mean": 0.6464506172839506,
|
| 464 |
-
"f1_sample_sd": 0.035742219952758225
|
| 465 |
-
},
|
| 466 |
-
{
|
| 467 |
-
"arm": "task_agnostic_base",
|
| 468 |
-
"label": "Loaded_Language",
|
| 469 |
-
"support": 45,
|
| 470 |
-
"precision_mean": 0.8649032907793016,
|
| 471 |
-
"precision_sample_sd": 0.046984312428427724,
|
| 472 |
-
"recall_mean": 0.9851851851851852,
|
| 473 |
-
"recall_sample_sd": 0.012830005981991702,
|
| 474 |
-
"f1_mean": 0.9208741623431562,
|
| 475 |
-
"f1_sample_sd": 0.03200926878737141
|
| 476 |
-
},
|
| 477 |
-
{
|
| 478 |
-
"arm": "task_agnostic_base",
|
| 479 |
-
"label": "Name_Calling,Labeling",
|
| 480 |
-
"support": 33,
|
| 481 |
-
"precision_mean": 0.7380952380952381,
|
| 482 |
-
"precision_sample_sd": 0.0047619047619048005,
|
| 483 |
-
"recall_mean": 0.9090909090909091,
|
| 484 |
-
"recall_sample_sd": 0.10925912955951485,
|
| 485 |
-
"f1_mean": 0.8125087983911513,
|
| 486 |
-
"f1_sample_sd": 0.04252971464057227
|
| 487 |
-
},
|
| 488 |
-
{
|
| 489 |
-
"arm": "task_agnostic_base",
|
| 490 |
-
"label": "Repetition",
|
| 491 |
-
"support": 23,
|
| 492 |
-
"precision_mean": 0.5196620583717357,
|
| 493 |
-
"precision_sample_sd": 0.021645909050194662,
|
| 494 |
-
"recall_mean": 0.7536231884057971,
|
| 495 |
-
"recall_sample_sd": 0.06641414050660639,
|
| 496 |
-
"f1_mean": 0.6147520781519571,
|
| 497 |
-
"f1_sample_sd": 0.03505927109416221
|
| 498 |
-
},
|
| 499 |
-
{
|
| 500 |
-
"arm": "task_agnostic_base",
|
| 501 |
-
"label": "Slogans",
|
| 502 |
-
"support": 14,
|
| 503 |
-
"precision_mean": 0.5161361626878869,
|
| 504 |
-
"precision_sample_sd": 0.20305677930281646,
|
| 505 |
-
"recall_mean": 0.7380952380952381,
|
| 506 |
-
"recall_sample_sd": 0.10910894511799614,
|
| 507 |
-
"f1_mean": 0.5834824090638044,
|
| 508 |
-
"f1_sample_sd": 0.09862686289220693
|
| 509 |
-
},
|
| 510 |
-
{
|
| 511 |
-
"arm": "task_agnostic_base",
|
| 512 |
-
"label": "Thought-terminating_Cliches",
|
| 513 |
-
"support": 7,
|
| 514 |
-
"precision_mean": 0.3492063492063492,
|
| 515 |
-
"precision_sample_sd": 0.143516994603374,
|
| 516 |
-
"recall_mean": 0.3333333333333333,
|
| 517 |
-
"recall_sample_sd": 0.08247860988423225,
|
| 518 |
-
"f1_mean": 0.319014319014319,
|
| 519 |
-
"f1_sample_sd": 0.04017590711117064
|
| 520 |
-
},
|
| 521 |
-
{
|
| 522 |
-
"arm": "task_agnostic_base",
|
| 523 |
-
"label": "Whataboutism,Straw_Men,Red_Herring",
|
| 524 |
-
"support": 10,
|
| 525 |
-
"precision_mean": 0.2761904761904762,
|
| 526 |
-
"precision_sample_sd": 0.05302632726504782,
|
| 527 |
-
"recall_mean": 0.5,
|
| 528 |
-
"recall_sample_sd": 0.2645751311064591,
|
| 529 |
-
"f1_mean": 0.330448343079922,
|
| 530 |
-
"f1_sample_sd": 0.021845164720973014
|
| 531 |
-
},
|
| 532 |
-
{
|
| 533 |
-
"arm": "structural_copy_control",
|
| 534 |
-
"label": "Appeal_to_Authority",
|
| 535 |
-
"support": 11,
|
| 536 |
-
"precision_mean": 0.26003086419753085,
|
| 537 |
-
"precision_sample_sd": 0.06281357095350094,
|
| 538 |
-
"recall_mean": 0.5757575757575758,
|
| 539 |
-
"recall_sample_sd": 0.262431940540739,
|
| 540 |
-
"f1_mean": 0.35477582846003897,
|
| 541 |
-
"f1_sample_sd": 0.11479479036518875
|
| 542 |
-
},
|
| 543 |
-
{
|
| 544 |
-
"arm": "structural_copy_control",
|
| 545 |
-
"label": "Appeal_to_fear-prejudice",
|
| 546 |
-
"support": 24,
|
| 547 |
-
"precision_mean": 0.5104813315339631,
|
| 548 |
-
"precision_sample_sd": 0.038427508996734555,
|
| 549 |
-
"recall_mean": 0.8611111111111112,
|
| 550 |
-
"recall_sample_sd": 0.024056261216234387,
|
| 551 |
-
"f1_mean": 0.6401745363877202,
|
| 552 |
-
"f1_sample_sd": 0.029305467347150463
|
| 553 |
-
},
|
| 554 |
-
{
|
| 555 |
-
"arm": "structural_copy_control",
|
| 556 |
-
"label": "Bandwagon,Reductio_ad_hitlerum",
|
| 557 |
-
"support": 7,
|
| 558 |
-
"precision_mean": 0.2876623376623376,
|
| 559 |
-
"precision_sample_sd": 0.061386823077618426,
|
| 560 |
-
"recall_mean": 0.8571428571428571,
|
| 561 |
-
"recall_sample_sd": 0.14285714285714285,
|
| 562 |
-
"f1_mean": 0.4306391777656145,
|
| 563 |
-
"f1_sample_sd": 0.08686392624022733
|
| 564 |
-
},
|
| 565 |
-
{
|
| 566 |
-
"arm": "structural_copy_control",
|
| 567 |
-
"label": "Black-and-White_Fallacy",
|
| 568 |
-
"support": 12,
|
| 569 |
-
"precision_mean": 0.31339031339031337,
|
| 570 |
-
"precision_sample_sd": 0.08301311842925053,
|
| 571 |
-
"recall_mean": 0.5277777777777778,
|
| 572 |
-
"recall_sample_sd": 0.2678791878053599,
|
| 573 |
-
"f1_mean": 0.38778427550357375,
|
| 574 |
-
"f1_sample_sd": 0.13069755090515112
|
| 575 |
-
},
|
| 576 |
-
{
|
| 577 |
-
"arm": "structural_copy_control",
|
| 578 |
-
"label": "Causal_Oversimplification",
|
| 579 |
-
"support": 18,
|
| 580 |
-
"precision_mean": 0.4464201043148412,
|
| 581 |
-
"precision_sample_sd": 0.013538446216471536,
|
| 582 |
-
"recall_mean": 0.9259259259259259,
|
| 583 |
-
"recall_sample_sd": 0.03207501495497923,
|
| 584 |
-
"f1_mean": 0.6023809523809524,
|
| 585 |
-
"f1_sample_sd": 0.018643640071557464
|
| 586 |
-
},
|
| 587 |
-
{
|
| 588 |
-
"arm": "structural_copy_control",
|
| 589 |
-
"label": "Doubt",
|
| 590 |
-
"support": 22,
|
| 591 |
-
"precision_mean": 0.4455314009661836,
|
| 592 |
-
"precision_sample_sd": 0.04067072694857813,
|
| 593 |
-
"recall_mean": 1.0,
|
| 594 |
-
"recall_sample_sd": 0.0,
|
| 595 |
-
"f1_mean": 0.615686274509804,
|
| 596 |
-
"f1_sample_sd": 0.039427200694482675
|
| 597 |
-
},
|
| 598 |
-
{
|
| 599 |
-
"arm": "structural_copy_control",
|
| 600 |
-
"label": "Exaggeration,Minimisation",
|
| 601 |
-
"support": 22,
|
| 602 |
-
"precision_mean": 0.4364176150420584,
|
| 603 |
-
"precision_sample_sd": 0.025093958106097894,
|
| 604 |
-
"recall_mean": 0.8787878787878788,
|
| 605 |
-
"recall_sample_sd": 0.05248638810814775,
|
| 606 |
-
"f1_mean": 0.5829803047194351,
|
| 607 |
-
"f1_sample_sd": 0.030899288356345502
|
| 608 |
-
},
|
| 609 |
-
{
|
| 610 |
-
"arm": "structural_copy_control",
|
| 611 |
-
"label": "Flag-Waving",
|
| 612 |
-
"support": 17,
|
| 613 |
-
"precision_mean": 0.3814814814814815,
|
| 614 |
-
"precision_sample_sd": 0.05010277503136551,
|
| 615 |
-
"recall_mean": 0.803921568627451,
|
| 616 |
-
"recall_sample_sd": 0.17970885078258195,
|
| 617 |
-
"f1_mean": 0.5138595287119652,
|
| 618 |
-
"f1_sample_sd": 0.06400931989326974
|
| 619 |
-
},
|
| 620 |
-
{
|
| 621 |
-
"arm": "structural_copy_control",
|
| 622 |
-
"label": "Loaded_Language",
|
| 623 |
-
"support": 45,
|
| 624 |
-
"precision_mean": 0.8492581841638446,
|
| 625 |
-
"precision_sample_sd": 0.016026591848331686,
|
| 626 |
-
"recall_mean": 1.0,
|
| 627 |
-
"recall_sample_sd": 0.0,
|
| 628 |
-
"f1_mean": 0.9184311025253588,
|
| 629 |
-
"f1_sample_sd": 0.009372233868132662
|
| 630 |
-
},
|
| 631 |
-
{
|
| 632 |
-
"arm": "structural_copy_control",
|
| 633 |
-
"label": "Name_Calling,Labeling",
|
| 634 |
-
"support": 33,
|
| 635 |
-
"precision_mean": 0.6375,
|
| 636 |
-
"precision_sample_sd": 0.03410563654946854,
|
| 637 |
-
"recall_mean": 0.9696969696969697,
|
| 638 |
-
"recall_sample_sd": 0.030303030303030276,
|
| 639 |
-
"f1_mean": 0.7685185185185185,
|
| 640 |
-
"f1_sample_sd": 0.02023900779105559
|
| 641 |
-
},
|
| 642 |
-
{
|
| 643 |
-
"arm": "structural_copy_control",
|
| 644 |
-
"label": "Repetition",
|
| 645 |
-
"support": 23,
|
| 646 |
-
"precision_mean": 0.44863933452168747,
|
| 647 |
-
"precision_sample_sd": 0.01520014564765896,
|
| 648 |
-
"recall_mean": 0.9420289855072463,
|
| 649 |
-
"recall_sample_sd": 0.06641414050660639,
|
| 650 |
-
"f1_mean": 0.607248835423033,
|
| 651 |
-
"f1_sample_sd": 0.019858627551810427
|
| 652 |
-
},
|
| 653 |
-
{
|
| 654 |
-
"arm": "structural_copy_control",
|
| 655 |
-
"label": "Slogans",
|
| 656 |
-
"support": 14,
|
| 657 |
-
"precision_mean": 0.3274318274318274,
|
| 658 |
-
"precision_sample_sd": 0.05434598572773684,
|
| 659 |
-
"recall_mean": 0.4523809523809524,
|
| 660 |
-
"recall_sample_sd": 0.1486904285332952,
|
| 661 |
-
"f1_mean": 0.3715828924162258,
|
| 662 |
-
"f1_sample_sd": 0.07106660236343577
|
| 663 |
-
},
|
| 664 |
-
{
|
| 665 |
-
"arm": "structural_copy_control",
|
| 666 |
-
"label": "Thought-terminating_Cliches",
|
| 667 |
-
"support": 7,
|
| 668 |
-
"precision_mean": 0.1142156862745098,
|
| 669 |
-
"precision_sample_sd": 0.012848374923608973,
|
| 670 |
-
"recall_mean": 0.38095238095238093,
|
| 671 |
-
"recall_sample_sd": 0.1649572197684645,
|
| 672 |
-
"f1_mean": 0.17331433998100665,
|
| 673 |
-
"f1_sample_sd": 0.029065880233243847
|
| 674 |
-
},
|
| 675 |
-
{
|
| 676 |
-
"arm": "structural_copy_control",
|
| 677 |
-
"label": "Whataboutism,Straw_Men,Red_Herring",
|
| 678 |
-
"support": 10,
|
| 679 |
-
"precision_mean": 0.19344598054275472,
|
| 680 |
-
"precision_sample_sd": 0.03373872471827226,
|
| 681 |
-
"recall_mean": 0.5666666666666667,
|
| 682 |
-
"recall_sample_sd": 0.2081665999466133,
|
| 683 |
-
"f1_mean": 0.28584083690299267,
|
| 684 |
-
"f1_sample_sd": 0.06078854830133676
|
| 685 |
-
},
|
| 686 |
-
{
|
| 687 |
-
"arm": "existing_specialized_separate_lineage",
|
| 688 |
-
"label": "Appeal_to_Authority",
|
| 689 |
-
"support": 11,
|
| 690 |
-
"precision_mean": 0.37836021505376344,
|
| 691 |
-
"precision_sample_sd": 0.10546365636105155,
|
| 692 |
-
"recall_mean": 0.6666666666666666,
|
| 693 |
-
"recall_sample_sd": 0.2287828616748712,
|
| 694 |
-
"f1_mean": 0.46885361552028215,
|
| 695 |
-
"f1_sample_sd": 0.09502747629890008
|
| 696 |
-
},
|
| 697 |
-
{
|
| 698 |
-
"arm": "existing_specialized_separate_lineage",
|
| 699 |
-
"label": "Appeal_to_fear-prejudice",
|
| 700 |
-
"support": 24,
|
| 701 |
-
"precision_mean": 0.525708635464733,
|
| 702 |
-
"precision_sample_sd": 0.02235198659615819,
|
| 703 |
-
"recall_mean": 0.8888888888888888,
|
| 704 |
-
"recall_sample_sd": 0.048112522432468774,
|
| 705 |
-
"f1_mean": 0.6599065351235072,
|
| 706 |
-
"f1_sample_sd": 0.015362381542320084
|
| 707 |
-
},
|
| 708 |
-
{
|
| 709 |
-
"arm": "existing_specialized_separate_lineage",
|
| 710 |
-
"label": "Bandwagon,Reductio_ad_hitlerum",
|
| 711 |
-
"support": 7,
|
| 712 |
-
"precision_mean": 0.44485144485144484,
|
| 713 |
-
"precision_sample_sd": 0.016487284575115283,
|
| 714 |
-
"recall_mean": 0.7619047619047619,
|
| 715 |
-
"recall_sample_sd": 0.1649572197684645,
|
| 716 |
-
"f1_mean": 0.5571428571428572,
|
| 717 |
-
"f1_sample_sd": 0.051507875363771265
|
| 718 |
-
},
|
| 719 |
-
{
|
| 720 |
-
"arm": "existing_specialized_separate_lineage",
|
| 721 |
-
"label": "Black-and-White_Fallacy",
|
| 722 |
-
"support": 12,
|
| 723 |
-
"precision_mean": 0.4662644544997486,
|
| 724 |
-
"precision_sample_sd": 0.004538274146363688,
|
| 725 |
-
"recall_mean": 0.5833333333333334,
|
| 726 |
-
"recall_sample_sd": 0.08333333333333331,
|
| 727 |
-
"f1_mean": 0.516747552149851,
|
| 728 |
-
"f1_sample_sd": 0.03589484965697345
|
| 729 |
-
},
|
| 730 |
-
{
|
| 731 |
-
"arm": "existing_specialized_separate_lineage",
|
| 732 |
-
"label": "Causal_Oversimplification",
|
| 733 |
-
"support": 18,
|
| 734 |
-
"precision_mean": 0.4472934472934473,
|
| 735 |
-
"precision_sample_sd": 0.013055771210700413,
|
| 736 |
-
"recall_mean": 0.8333333333333333,
|
| 737 |
-
"recall_sample_sd": 0.1469861839480328,
|
| 738 |
-
"f1_mean": 0.5781794553724379,
|
| 739 |
-
"f1_sample_sd": 0.028407563015813122
|
| 740 |
-
},
|
| 741 |
-
{
|
| 742 |
-
"arm": "existing_specialized_separate_lineage",
|
| 743 |
-
"label": "Doubt",
|
| 744 |
-
"support": 22,
|
| 745 |
-
"precision_mean": 0.5045661001171619,
|
| 746 |
-
"precision_sample_sd": 0.023581174969843577,
|
| 747 |
-
"recall_mean": 1.0,
|
| 748 |
-
"recall_sample_sd": 0.0,
|
| 749 |
-
"f1_mean": 0.6704939668174963,
|
| 750 |
-
"f1_sample_sd": 0.020973132363659372
|
| 751 |
-
},
|
| 752 |
-
{
|
| 753 |
-
"arm": "existing_specialized_separate_lineage",
|
| 754 |
-
"label": "Exaggeration,Minimisation",
|
| 755 |
-
"support": 22,
|
| 756 |
-
"precision_mean": 0.44589793281653745,
|
| 757 |
-
"precision_sample_sd": 0.010987847993006481,
|
| 758 |
-
"recall_mean": 0.9393939393939394,
|
| 759 |
-
"recall_sample_sd": 0.06943296507508849,
|
| 760 |
-
"f1_mean": 0.6043956043956044,
|
| 761 |
-
"f1_sample_sd": 0.022305256187015846
|
| 762 |
-
},
|
| 763 |
-
{
|
| 764 |
-
"arm": "existing_specialized_separate_lineage",
|
| 765 |
-
"label": "Flag-Waving",
|
| 766 |
-
"support": 17,
|
| 767 |
-
"precision_mean": 0.5576388888888889,
|
| 768 |
-
"precision_sample_sd": 0.03705367103320583,
|
| 769 |
-
"recall_mean": 0.8823529411764706,
|
| 770 |
-
"recall_sample_sd": 0.11764705882352944,
|
| 771 |
-
"f1_mean": 0.680769868923179,
|
| 772 |
-
"f1_sample_sd": 0.04164658769406446
|
| 773 |
-
},
|
| 774 |
-
{
|
| 775 |
-
"arm": "existing_specialized_separate_lineage",
|
| 776 |
-
"label": "Loaded_Language",
|
| 777 |
-
"support": 45,
|
| 778 |
-
"precision_mean": 0.8387650085763294,
|
| 779 |
-
"precision_sample_sd": 0.01782556577257849,
|
| 780 |
-
"recall_mean": 1.0,
|
| 781 |
-
"recall_sample_sd": 0.0,
|
| 782 |
-
"f1_mean": 0.9122448979591837,
|
| 783 |
-
"f1_sample_sd": 0.01060439269940129
|
| 784 |
-
},
|
| 785 |
-
{
|
| 786 |
-
"arm": "existing_specialized_separate_lineage",
|
| 787 |
-
"label": "Name_Calling,Labeling",
|
| 788 |
-
"support": 33,
|
| 789 |
-
"precision_mean": 0.6957104190221673,
|
| 790 |
-
"precision_sample_sd": 0.019201490809269257,
|
| 791 |
-
"recall_mean": 0.9696969696969697,
|
| 792 |
-
"recall_sample_sd": 0.030303030303030276,
|
| 793 |
-
"f1_mean": 0.8101049442821594,
|
| 794 |
-
"f1_sample_sd": 0.022092733328802537
|
| 795 |
-
},
|
| 796 |
-
{
|
| 797 |
-
"arm": "existing_specialized_separate_lineage",
|
| 798 |
-
"label": "Repetition",
|
| 799 |
-
"support": 23,
|
| 800 |
-
"precision_mean": 0.4781557067271353,
|
| 801 |
-
"precision_sample_sd": 0.009897987322673213,
|
| 802 |
-
"recall_mean": 0.9420289855072463,
|
| 803 |
-
"recall_sample_sd": 0.06641414050660639,
|
| 804 |
-
"f1_mean": 0.6337774426009719,
|
| 805 |
-
"f1_sample_sd": 0.016444118333485254
|
| 806 |
-
},
|
| 807 |
-
{
|
| 808 |
-
"arm": "existing_specialized_separate_lineage",
|
| 809 |
-
"label": "Slogans",
|
| 810 |
-
"support": 14,
|
| 811 |
-
"precision_mean": 0.4334215167548501,
|
| 812 |
-
"precision_sample_sd": 0.05811581861282867,
|
| 813 |
-
"recall_mean": 0.6904761904761905,
|
| 814 |
-
"recall_sample_sd": 0.1649572197684645,
|
| 815 |
-
"f1_mean": 0.5201316298877274,
|
| 816 |
-
"f1_sample_sd": 0.018567913895656114
|
| 817 |
-
},
|
| 818 |
-
{
|
| 819 |
-
"arm": "existing_specialized_separate_lineage",
|
| 820 |
-
"label": "Thought-terminating_Cliches",
|
| 821 |
-
"support": 7,
|
| 822 |
-
"precision_mean": 0.18446115288220552,
|
| 823 |
-
"precision_sample_sd": 0.03641251252864963,
|
| 824 |
-
"recall_mean": 0.6666666666666666,
|
| 825 |
-
"recall_sample_sd": 0.1649572197684645,
|
| 826 |
-
"f1_mean": 0.2868626868626869,
|
| 827 |
-
"f1_sample_sd": 0.0511620864429468
|
| 828 |
-
},
|
| 829 |
-
{
|
| 830 |
-
"arm": "existing_specialized_separate_lineage",
|
| 831 |
-
"label": "Whataboutism,Straw_Men,Red_Herring",
|
| 832 |
-
"support": 10,
|
| 833 |
-
"precision_mean": 0.2777777777777778,
|
| 834 |
-
"precision_sample_sd": 0.0481125224324688,
|
| 835 |
-
"recall_mean": 0.4,
|
| 836 |
-
"recall_sample_sd": 0.0,
|
| 837 |
-
"f1_mean": 0.32634032634032634,
|
| 838 |
-
"f1_sample_sd": 0.03229931575886014
|
| 839 |
-
}
|
| 840 |
-
],
|
| 841 |
-
"compression_ladder": [
|
| 842 |
-
{
|
| 843 |
-
"name": "24L upstream text teacher",
|
| 844 |
-
"layers": 24,
|
| 845 |
-
"parameters": 752393024,
|
| 846 |
-
"weight_bytes": 1504825632,
|
| 847 |
-
"weight_mib": 1435.1135559082031,
|
| 848 |
-
"weight_sha256": "30783da4651259c26b8a204821ea426b9526431e0c14da0d1c6fa284e7414e5e",
|
| 849 |
-
"kd_stage": "",
|
| 850 |
-
"mean_total_loss": "",
|
| 851 |
-
"mean_interface_loss": "",
|
| 852 |
-
"mean_final_loss": ""
|
| 853 |
-
},
|
| 854 |
-
{
|
| 855 |
-
"name": "8L task-free KD",
|
| 856 |
-
"layers": 8,
|
| 857 |
-
"parameters": 420318144,
|
| 858 |
-
"weight_bytes": 840647880,
|
| 859 |
-
"weight_mib": 801.7042922973633,
|
| 860 |
-
"weight_sha256": "086fbaf9a4838ecde56cd3541c9c3212b22d0c5ee745cf2c1c5e0bb31e0ec22c",
|
| 861 |
-
"kd_stage": "24to8",
|
| 862 |
-
"mean_total_loss": 0.17622653172702485,
|
| 863 |
-
"mean_interface_loss": 0.21562811763578793,
|
| 864 |
-
"mean_final_loss": 0.16626427527853593
|
| 865 |
-
},
|
| 866 |
-
{
|
| 867 |
-
"name": "6L task-free KD",
|
| 868 |
-
"layers": 6,
|
| 869 |
-
"parameters": 377207424,
|
| 870 |
-
"weight_bytes": 754423424,
|
| 871 |
-
"weight_mib": 719.4742431640625,
|
| 872 |
-
"weight_sha256": "492749c20f77a42f6bc993f0927a82217b94f872b7eacf25612f7520c9bc28ce",
|
| 873 |
-
"kd_stage": "8to6",
|
| 874 |
-
"mean_total_loss": 0.03382104352249371,
|
| 875 |
-
"mean_interface_loss": 0.04486069045515251,
|
| 876 |
-
"mean_final_loss": 0.031147060785087888
|
| 877 |
-
},
|
| 878 |
-
{
|
| 879 |
-
"name": "4L task-free KD root",
|
| 880 |
-
"layers": 4,
|
| 881 |
-
"parameters": 334096704,
|
| 882 |
-
"weight_bytes": 668198976,
|
| 883 |
-
"weight_mib": 637.2442016601562,
|
| 884 |
-
"weight_sha256": "2732c616772fe320cdea228ab4554981418b1b2bf615c4183fb1ac8e6e2168d3",
|
| 885 |
-
"kd_stage": "6to4",
|
| 886 |
-
"mean_total_loss": 0.047020394468859195,
|
| 887 |
-
"mean_interface_loss": 0.07048421185527332,
|
| 888 |
-
"mean_final_loss": 0.043119939065377366
|
| 889 |
-
}
|
| 890 |
-
]
|
| 891 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@@ -1,891 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"schema": "standalone4l-complete-transfer-benchmark-v2",
|
| 3 |
-
"status": "complete_from_frozen_reports",
|
| 4 |
-
"claim_scope": "Exploratory fresh-head transfer evidence on one English task and a previously opened 55-article SemEval-derived test split.",
|
| 5 |
-
"headless_root_requires_adaptation": true,
|
| 6 |
-
"test_previously_opened": true,
|
| 7 |
-
"test_articles": 55,
|
| 8 |
-
"test_windows": 434,
|
| 9 |
-
"seeds": [
|
| 10 |
-
41,
|
| 11 |
-
42,
|
| 12 |
-
43
|
| 13 |
-
],
|
| 14 |
-
"three_seed_sd_is_not_confidence_interval": true,
|
| 15 |
-
"multilingual_downstream_evaluation_performed": false,
|
| 16 |
-
"upstream_multilingual_metadata": {
|
| 17 |
-
"source": "Qwen/Qwen3.5-0.8B official model card",
|
| 18 |
-
"languages_and_dialects_claimed_upstream": 201,
|
| 19 |
-
"scope": "inherited tokenizer and architecture coverage only"
|
| 20 |
-
},
|
| 21 |
-
"protocol": {
|
| 22 |
-
"epochs": 5,
|
| 23 |
-
"primary_epoch": 5,
|
| 24 |
-
"early_stopping": false,
|
| 25 |
-
"max_length": 256,
|
| 26 |
-
"stride": 128,
|
| 27 |
-
"window_target": "span-aligned technique labels",
|
| 28 |
-
"article_aggregation": "labelwise maximum probability",
|
| 29 |
-
"micro_batch_size": 1,
|
| 30 |
-
"gradient_accumulation": 32,
|
| 31 |
-
"effective_batch_size": 32,
|
| 32 |
-
"learning_rate": 2e-05,
|
| 33 |
-
"weight_decay": 0.01,
|
| 34 |
-
"gradient_clip": 1.0,
|
| 35 |
-
"optimizer": "AdamW",
|
| 36 |
-
"dtype": "bfloat16",
|
| 37 |
-
"thresholds": [
|
| 38 |
-
0.05,
|
| 39 |
-
0.1,
|
| 40 |
-
0.15,
|
| 41 |
-
0.2,
|
| 42 |
-
0.25,
|
| 43 |
-
0.3,
|
| 44 |
-
0.35,
|
| 45 |
-
0.4,
|
| 46 |
-
0.45,
|
| 47 |
-
0.5,
|
| 48 |
-
0.55,
|
| 49 |
-
0.6,
|
| 50 |
-
0.65,
|
| 51 |
-
0.7,
|
| 52 |
-
0.75,
|
| 53 |
-
0.8,
|
| 54 |
-
0.85,
|
| 55 |
-
0.9
|
| 56 |
-
],
|
| 57 |
-
"threshold_selection": "epoch-5 calibration macro-F1, then micro-F1"
|
| 58 |
-
},
|
| 59 |
-
"arms": {
|
| 60 |
-
"task_agnostic_base": {
|
| 61 |
-
"display": "Task-agnostic KD base",
|
| 62 |
-
"lineage": "root base plus a fresh 14-label transfer head",
|
| 63 |
-
"metrics": {
|
| 64 |
-
"macro_f1": {
|
| 65 |
-
"values": [
|
| 66 |
-
0.5883093293517536,
|
| 67 |
-
0.6028379036218096,
|
| 68 |
-
0.5712816197954101
|
| 69 |
-
],
|
| 70 |
-
"mean": 0.5874762842563245,
|
| 71 |
-
"sample_sd": 0.01579462678670608,
|
| 72 |
-
"n": 3
|
| 73 |
-
},
|
| 74 |
-
"micro_f1": {
|
| 75 |
-
"values": [
|
| 76 |
-
0.6468085106382978,
|
| 77 |
-
0.6655172413793103,
|
| 78 |
-
0.6465256797583082
|
| 79 |
-
],
|
| 80 |
-
"mean": 0.6529504772586388,
|
| 81 |
-
"sample_sd": 0.010884055708931178,
|
| 82 |
-
"n": 3
|
| 83 |
-
},
|
| 84 |
-
"exact_match": {
|
| 85 |
-
"values": [
|
| 86 |
-
0.05454545454545454,
|
| 87 |
-
0.03636363636363636,
|
| 88 |
-
0.0
|
| 89 |
-
],
|
| 90 |
-
"mean": 0.0303030303030303,
|
| 91 |
-
"sample_sd": 0.027773186030035393,
|
| 92 |
-
"n": 3
|
| 93 |
-
},
|
| 94 |
-
"predicted_positive_rate": {
|
| 95 |
-
"values": [
|
| 96 |
-
0.5714285714285714,
|
| 97 |
-
0.4090909090909091,
|
| 98 |
-
0.5155844155844156
|
| 99 |
-
],
|
| 100 |
-
"mean": 0.4987012987012987,
|
| 101 |
-
"sample_sd": 0.08247520160679181,
|
| 102 |
-
"n": 3
|
| 103 |
-
}
|
| 104 |
-
},
|
| 105 |
-
"resources": {
|
| 106 |
-
"parameters": {
|
| 107 |
-
"values": [
|
| 108 |
-
334111040.0,
|
| 109 |
-
334111040.0,
|
| 110 |
-
334111040.0
|
| 111 |
-
],
|
| 112 |
-
"mean": 334111040.0,
|
| 113 |
-
"sample_sd": 0.0,
|
| 114 |
-
"n": 3
|
| 115 |
-
},
|
| 116 |
-
"peak_cuda_allocated_bytes": {
|
| 117 |
-
"values": [
|
| 118 |
-
3380975616.0,
|
| 119 |
-
3380975616.0,
|
| 120 |
-
3380975616.0
|
| 121 |
-
],
|
| 122 |
-
"mean": 3380975616.0,
|
| 123 |
-
"sample_sd": 0.0,
|
| 124 |
-
"n": 3
|
| 125 |
-
},
|
| 126 |
-
"peak_cuda_reserved_bytes": {
|
| 127 |
-
"values": [
|
| 128 |
-
3519021056.0,
|
| 129 |
-
3519021056.0,
|
| 130 |
-
3510632448.0
|
| 131 |
-
],
|
| 132 |
-
"mean": 3516224853.3333335,
|
| 133 |
-
"sample_sd": 4843165.086926248,
|
| 134 |
-
"n": 3
|
| 135 |
-
},
|
| 136 |
-
"total_wall_seconds": {
|
| 137 |
-
"values": [
|
| 138 |
-
1583.0297977919981,
|
| 139 |
-
1588.617994428001,
|
| 140 |
-
1603.4310500120046
|
| 141 |
-
],
|
| 142 |
-
"mean": 1591.692947410668,
|
| 143 |
-
"sample_sd": 10.54249851412311,
|
| 144 |
-
"n": 3
|
| 145 |
-
},
|
| 146 |
-
"test_seconds_per_article": {
|
| 147 |
-
"values": [
|
| 148 |
-
0.08164824154545468,
|
| 149 |
-
0.07871871667289683,
|
| 150 |
-
0.08355205721825107
|
| 151 |
-
],
|
| 152 |
-
"mean": 0.08130633847886752,
|
| 153 |
-
"sample_sd": 0.0024347419754560395,
|
| 154 |
-
"n": 3
|
| 155 |
-
}
|
| 156 |
-
},
|
| 157 |
-
"thresholds": [
|
| 158 |
-
0.15,
|
| 159 |
-
0.3,
|
| 160 |
-
0.1
|
| 161 |
-
]
|
| 162 |
-
},
|
| 163 |
-
"structural_copy_control": {
|
| 164 |
-
"display": "Structural copy, no task-free KD",
|
| 165 |
-
"lineage": "same four-layer structure without task-free KD",
|
| 166 |
-
"metrics": {
|
| 167 |
-
"macro_f1": {
|
| 168 |
-
"values": [
|
| 169 |
-
0.5169424579262757,
|
| 170 |
-
0.5349405779540175,
|
| 171 |
-
0.5023778364496152
|
| 172 |
-
],
|
| 173 |
-
"mean": 0.5180869574433028,
|
| 174 |
-
"sample_sd": 0.016311512588628545,
|
| 175 |
-
"n": 3
|
| 176 |
-
},
|
| 177 |
-
"micro_f1": {
|
| 178 |
-
"values": [
|
| 179 |
-
0.5773955773955775,
|
| 180 |
-
0.6083445491251682,
|
| 181 |
-
0.585635359116022
|
| 182 |
-
],
|
| 183 |
-
"mean": 0.5904584952122559,
|
| 184 |
-
"sample_sd": 0.01602830914762526,
|
| 185 |
-
"n": 3
|
| 186 |
-
},
|
| 187 |
-
"exact_match": {
|
| 188 |
-
"values": [
|
| 189 |
-
0.0,
|
| 190 |
-
0.0,
|
| 191 |
-
0.0
|
| 192 |
-
],
|
| 193 |
-
"mean": 0.0,
|
| 194 |
-
"sample_sd": 0.0,
|
| 195 |
-
"n": 3
|
| 196 |
-
},
|
| 197 |
-
"predicted_positive_rate": {
|
| 198 |
-
"values": [
|
| 199 |
-
0.712987012987013,
|
| 200 |
-
0.6207792207792208,
|
| 201 |
-
0.5961038961038961
|
| 202 |
-
],
|
| 203 |
-
"mean": 0.6432900432900432,
|
| 204 |
-
"sample_sd": 0.06160737456395461,
|
| 205 |
-
"n": 3
|
| 206 |
-
}
|
| 207 |
-
},
|
| 208 |
-
"resources": {
|
| 209 |
-
"parameters": {
|
| 210 |
-
"values": [
|
| 211 |
-
334111040.0,
|
| 212 |
-
334111040.0,
|
| 213 |
-
334111040.0
|
| 214 |
-
],
|
| 215 |
-
"mean": 334111040.0,
|
| 216 |
-
"sample_sd": 0.0,
|
| 217 |
-
"n": 3
|
| 218 |
-
},
|
| 219 |
-
"peak_cuda_allocated_bytes": {
|
| 220 |
-
"values": [
|
| 221 |
-
3380975616.0,
|
| 222 |
-
3380975616.0,
|
| 223 |
-
3380975616.0
|
| 224 |
-
],
|
| 225 |
-
"mean": 3380975616.0,
|
| 226 |
-
"sample_sd": 0.0,
|
| 227 |
-
"n": 3
|
| 228 |
-
},
|
| 229 |
-
"peak_cuda_reserved_bytes": {
|
| 230 |
-
"values": [
|
| 231 |
-
3519021056.0,
|
| 232 |
-
3519021056.0,
|
| 233 |
-
3510632448.0
|
| 234 |
-
],
|
| 235 |
-
"mean": 3516224853.3333335,
|
| 236 |
-
"sample_sd": 4843165.086926248,
|
| 237 |
-
"n": 3
|
| 238 |
-
},
|
| 239 |
-
"total_wall_seconds": {
|
| 240 |
-
"values": [
|
| 241 |
-
1575.5905925639963,
|
| 242 |
-
1623.5639530190092,
|
| 243 |
-
1629.195114912989
|
| 244 |
-
],
|
| 245 |
-
"mean": 1609.4498868319981,
|
| 246 |
-
"sample_sd": 29.457874198835384,
|
| 247 |
-
"n": 3
|
| 248 |
-
},
|
| 249 |
-
"test_seconds_per_article": {
|
| 250 |
-
"values": [
|
| 251 |
-
0.07877608058201573,
|
| 252 |
-
0.08307423260041767,
|
| 253 |
-
0.08163602005436339
|
| 254 |
-
],
|
| 255 |
-
"mean": 0.08116211107893226,
|
| 256 |
-
"sample_sd": 0.002187914527825257,
|
| 257 |
-
"n": 3
|
| 258 |
-
}
|
| 259 |
-
},
|
| 260 |
-
"thresholds": [
|
| 261 |
-
0.05,
|
| 262 |
-
0.2,
|
| 263 |
-
0.2
|
| 264 |
-
]
|
| 265 |
-
},
|
| 266 |
-
"existing_specialized_separate_lineage": {
|
| 267 |
-
"display": "Existing specialized 4L, separate lineage",
|
| 268 |
-
"lineage": "historical task-specific shrink/distillation lineage",
|
| 269 |
-
"metrics": {
|
| 270 |
-
"macro_f1": {
|
| 271 |
-
"values": [
|
| 272 |
-
0.5832848452288824,
|
| 273 |
-
0.5958405181160762,
|
| 274 |
-
0.5835785045218137
|
| 275 |
-
],
|
| 276 |
-
"mean": 0.5875679559555907,
|
| 277 |
-
"sample_sd": 0.007165753447433348,
|
| 278 |
-
"n": 3
|
| 279 |
-
},
|
| 280 |
-
"micro_f1": {
|
| 281 |
-
"values": [
|
| 282 |
-
0.6494252873563218,
|
| 283 |
-
0.6458616010854816,
|
| 284 |
-
0.6486486486486486
|
| 285 |
-
],
|
| 286 |
-
"mean": 0.647978512363484,
|
| 287 |
-
"sample_sd": 0.0018739735696546695,
|
| 288 |
-
"n": 3
|
| 289 |
-
},
|
| 290 |
-
"exact_match": {
|
| 291 |
-
"values": [
|
| 292 |
-
0.0,
|
| 293 |
-
0.0,
|
| 294 |
-
0.01818181818181818
|
| 295 |
-
],
|
| 296 |
-
"mean": 0.006060606060606061,
|
| 297 |
-
"sample_sd": 0.01049727762162956,
|
| 298 |
-
"n": 3
|
| 299 |
-
},
|
| 300 |
-
"predicted_positive_rate": {
|
| 301 |
-
"values": [
|
| 302 |
-
0.5597402597402598,
|
| 303 |
-
0.612987012987013,
|
| 304 |
-
0.5688311688311688
|
| 305 |
-
],
|
| 306 |
-
"mean": 0.5805194805194805,
|
| 307 |
-
"sample_sd": 0.02848274311618351,
|
| 308 |
-
"n": 3
|
| 309 |
-
}
|
| 310 |
-
},
|
| 311 |
-
"resources": {
|
| 312 |
-
"parameters": {
|
| 313 |
-
"values": [
|
| 314 |
-
210903360.0,
|
| 315 |
-
210903360.0,
|
| 316 |
-
210903360.0
|
| 317 |
-
],
|
| 318 |
-
"mean": 210903360.0,
|
| 319 |
-
"sample_sd": 0.0,
|
| 320 |
-
"n": 3
|
| 321 |
-
},
|
| 322 |
-
"peak_cuda_allocated_bytes": {
|
| 323 |
-
"values": [
|
| 324 |
-
2143655936.0,
|
| 325 |
-
2143655936.0,
|
| 326 |
-
2143655936.0
|
| 327 |
-
],
|
| 328 |
-
"mean": 2143655936.0,
|
| 329 |
-
"sample_sd": 0.0,
|
| 330 |
-
"n": 3
|
| 331 |
-
},
|
| 332 |
-
"peak_cuda_reserved_bytes": {
|
| 333 |
-
"values": [
|
| 334 |
-
2281701376.0,
|
| 335 |
-
2281701376.0,
|
| 336 |
-
2273312768.0
|
| 337 |
-
],
|
| 338 |
-
"mean": 2278905173.3333335,
|
| 339 |
-
"sample_sd": 4843165.086926248,
|
| 340 |
-
"n": 3
|
| 341 |
-
},
|
| 342 |
-
"total_wall_seconds": {
|
| 343 |
-
"values": [
|
| 344 |
-
1634.488557530014,
|
| 345 |
-
1652.9397718419787,
|
| 346 |
-
1643.488202479988
|
| 347 |
-
],
|
| 348 |
-
"mean": 1643.6388439506602,
|
| 349 |
-
"sample_sd": 9.226529522850198,
|
| 350 |
-
"n": 3
|
| 351 |
-
},
|
| 352 |
-
"test_seconds_per_article": {
|
| 353 |
-
"values": [
|
| 354 |
-
0.0847959352360869,
|
| 355 |
-
0.08598086021798239,
|
| 356 |
-
0.08552987641781908
|
| 357 |
-
],
|
| 358 |
-
"mean": 0.08543555729062946,
|
| 359 |
-
"sample_sd": 0.0005980667826383295,
|
| 360 |
-
"n": 3
|
| 361 |
-
}
|
| 362 |
-
},
|
| 363 |
-
"thresholds": [
|
| 364 |
-
0.1,
|
| 365 |
-
0.05,
|
| 366 |
-
0.05
|
| 367 |
-
]
|
| 368 |
-
}
|
| 369 |
-
},
|
| 370 |
-
"same_seed_macro_f1_deltas_base_minus_control": [
|
| 371 |
-
0.07136687142547793,
|
| 372 |
-
0.06789732566779205,
|
| 373 |
-
0.06890378334579494
|
| 374 |
-
],
|
| 375 |
-
"mean_macro_f1_delta_base_minus_control": 0.06938932681302164,
|
| 376 |
-
"publication_gate_passed": true,
|
| 377 |
-
"per_label_summary": [
|
| 378 |
-
{
|
| 379 |
-
"arm": "task_agnostic_base",
|
| 380 |
-
"label": "Appeal_to_Authority",
|
| 381 |
-
"support": 11,
|
| 382 |
-
"precision_mean": 0.3958614484930274,
|
| 383 |
-
"precision_sample_sd": 0.1238064457010119,
|
| 384 |
-
"recall_mean": 0.5757575757575757,
|
| 385 |
-
"recall_sample_sd": 0.05248638810814781,
|
| 386 |
-
"f1_mean": 0.4657088122605364,
|
| 387 |
-
"f1_sample_sd": 0.1020990126979069
|
| 388 |
-
},
|
| 389 |
-
{
|
| 390 |
-
"arm": "task_agnostic_base",
|
| 391 |
-
"label": "Appeal_to_fear-prejudice",
|
| 392 |
-
"support": 24,
|
| 393 |
-
"precision_mean": 0.5540244550770866,
|
| 394 |
-
"precision_sample_sd": 0.02525673990402998,
|
| 395 |
-
"recall_mean": 0.7361111111111112,
|
| 396 |
-
"recall_sample_sd": 0.13393959390267993,
|
| 397 |
-
"f1_mean": 0.6277521761392729,
|
| 398 |
-
"f1_sample_sd": 0.04994879016183189
|
| 399 |
-
},
|
| 400 |
-
{
|
| 401 |
-
"arm": "task_agnostic_base",
|
| 402 |
-
"label": "Bandwagon,Reductio_ad_hitlerum",
|
| 403 |
-
"support": 7,
|
| 404 |
-
"precision_mean": 0.4851851851851852,
|
| 405 |
-
"precision_sample_sd": 0.07882887649552875,
|
| 406 |
-
"recall_mean": 0.8095238095238095,
|
| 407 |
-
"recall_sample_sd": 0.08247860988423221,
|
| 408 |
-
"f1_mean": 0.6006778309409889,
|
| 409 |
-
"f1_sample_sd": 0.04793776258918466
|
| 410 |
-
},
|
| 411 |
-
{
|
| 412 |
-
"arm": "task_agnostic_base",
|
| 413 |
-
"label": "Black-and-White_Fallacy",
|
| 414 |
-
"support": 12,
|
| 415 |
-
"precision_mean": 0.3667929292929293,
|
| 416 |
-
"precision_sample_sd": 0.07670649392711647,
|
| 417 |
-
"recall_mean": 0.47222222222222227,
|
| 418 |
-
"recall_sample_sd": 0.09622504486493764,
|
| 419 |
-
"f1_mean": 0.4053892966936445,
|
| 420 |
-
"f1_sample_sd": 0.04211369218048746
|
| 421 |
-
},
|
| 422 |
-
{
|
| 423 |
-
"arm": "task_agnostic_base",
|
| 424 |
-
"label": "Causal_Oversimplification",
|
| 425 |
-
"support": 18,
|
| 426 |
-
"precision_mean": 0.5428824049513704,
|
| 427 |
-
"precision_sample_sd": 0.039216350192122156,
|
| 428 |
-
"recall_mean": 0.7962962962962963,
|
| 429 |
-
"recall_sample_sd": 0.11564811108145183,
|
| 430 |
-
"f1_mean": 0.6446992723588468,
|
| 431 |
-
"f1_sample_sd": 0.0634561308339175
|
| 432 |
-
},
|
| 433 |
-
{
|
| 434 |
-
"arm": "task_agnostic_base",
|
| 435 |
-
"label": "Doubt",
|
| 436 |
-
"support": 22,
|
| 437 |
-
"precision_mean": 0.5223832528180354,
|
| 438 |
-
"precision_sample_sd": 0.06742780050556177,
|
| 439 |
-
"recall_mean": 0.9848484848484849,
|
| 440 |
-
"recall_sample_sd": 0.026243194054073875,
|
| 441 |
-
"f1_mean": 0.6802057822343391,
|
| 442 |
-
"f1_sample_sd": 0.04928561754686331
|
| 443 |
-
},
|
| 444 |
-
{
|
| 445 |
-
"arm": "task_agnostic_base",
|
| 446 |
-
"label": "Exaggeration,Minimisation",
|
| 447 |
-
"support": 22,
|
| 448 |
-
"precision_mean": 0.4620811287477954,
|
| 449 |
-
"precision_sample_sd": 0.016801198309751,
|
| 450 |
-
"recall_mean": 0.7727272727272727,
|
| 451 |
-
"recall_sample_sd": 0.15745916432444337,
|
| 452 |
-
"f1_mean": 0.5727040816326531,
|
| 453 |
-
"f1_sample_sd": 0.03645259990419198
|
| 454 |
-
},
|
| 455 |
-
{
|
| 456 |
-
"arm": "task_agnostic_base",
|
| 457 |
-
"label": "Flag-Waving",
|
| 458 |
-
"support": 17,
|
| 459 |
-
"precision_mean": 0.5642642642642642,
|
| 460 |
-
"precision_sample_sd": 0.1478145677656334,
|
| 461 |
-
"recall_mean": 0.8235294117647058,
|
| 462 |
-
"recall_sample_sd": 0.1764705882352941,
|
| 463 |
-
"f1_mean": 0.6464506172839506,
|
| 464 |
-
"f1_sample_sd": 0.035742219952758225
|
| 465 |
-
},
|
| 466 |
-
{
|
| 467 |
-
"arm": "task_agnostic_base",
|
| 468 |
-
"label": "Loaded_Language",
|
| 469 |
-
"support": 45,
|
| 470 |
-
"precision_mean": 0.8649032907793016,
|
| 471 |
-
"precision_sample_sd": 0.046984312428427724,
|
| 472 |
-
"recall_mean": 0.9851851851851852,
|
| 473 |
-
"recall_sample_sd": 0.012830005981991702,
|
| 474 |
-
"f1_mean": 0.9208741623431562,
|
| 475 |
-
"f1_sample_sd": 0.03200926878737141
|
| 476 |
-
},
|
| 477 |
-
{
|
| 478 |
-
"arm": "task_agnostic_base",
|
| 479 |
-
"label": "Name_Calling,Labeling",
|
| 480 |
-
"support": 33,
|
| 481 |
-
"precision_mean": 0.7380952380952381,
|
| 482 |
-
"precision_sample_sd": 0.0047619047619048005,
|
| 483 |
-
"recall_mean": 0.9090909090909091,
|
| 484 |
-
"recall_sample_sd": 0.10925912955951485,
|
| 485 |
-
"f1_mean": 0.8125087983911513,
|
| 486 |
-
"f1_sample_sd": 0.04252971464057227
|
| 487 |
-
},
|
| 488 |
-
{
|
| 489 |
-
"arm": "task_agnostic_base",
|
| 490 |
-
"label": "Repetition",
|
| 491 |
-
"support": 23,
|
| 492 |
-
"precision_mean": 0.5196620583717357,
|
| 493 |
-
"precision_sample_sd": 0.021645909050194662,
|
| 494 |
-
"recall_mean": 0.7536231884057971,
|
| 495 |
-
"recall_sample_sd": 0.06641414050660639,
|
| 496 |
-
"f1_mean": 0.6147520781519571,
|
| 497 |
-
"f1_sample_sd": 0.03505927109416221
|
| 498 |
-
},
|
| 499 |
-
{
|
| 500 |
-
"arm": "task_agnostic_base",
|
| 501 |
-
"label": "Slogans",
|
| 502 |
-
"support": 14,
|
| 503 |
-
"precision_mean": 0.5161361626878869,
|
| 504 |
-
"precision_sample_sd": 0.20305677930281646,
|
| 505 |
-
"recall_mean": 0.7380952380952381,
|
| 506 |
-
"recall_sample_sd": 0.10910894511799614,
|
| 507 |
-
"f1_mean": 0.5834824090638044,
|
| 508 |
-
"f1_sample_sd": 0.09862686289220693
|
| 509 |
-
},
|
| 510 |
-
{
|
| 511 |
-
"arm": "task_agnostic_base",
|
| 512 |
-
"label": "Thought-terminating_Cliches",
|
| 513 |
-
"support": 7,
|
| 514 |
-
"precision_mean": 0.3492063492063492,
|
| 515 |
-
"precision_sample_sd": 0.143516994603374,
|
| 516 |
-
"recall_mean": 0.3333333333333333,
|
| 517 |
-
"recall_sample_sd": 0.08247860988423225,
|
| 518 |
-
"f1_mean": 0.319014319014319,
|
| 519 |
-
"f1_sample_sd": 0.04017590711117064
|
| 520 |
-
},
|
| 521 |
-
{
|
| 522 |
-
"arm": "task_agnostic_base",
|
| 523 |
-
"label": "Whataboutism,Straw_Men,Red_Herring",
|
| 524 |
-
"support": 10,
|
| 525 |
-
"precision_mean": 0.2761904761904762,
|
| 526 |
-
"precision_sample_sd": 0.05302632726504782,
|
| 527 |
-
"recall_mean": 0.5,
|
| 528 |
-
"recall_sample_sd": 0.2645751311064591,
|
| 529 |
-
"f1_mean": 0.330448343079922,
|
| 530 |
-
"f1_sample_sd": 0.021845164720973014
|
| 531 |
-
},
|
| 532 |
-
{
|
| 533 |
-
"arm": "structural_copy_control",
|
| 534 |
-
"label": "Appeal_to_Authority",
|
| 535 |
-
"support": 11,
|
| 536 |
-
"precision_mean": 0.26003086419753085,
|
| 537 |
-
"precision_sample_sd": 0.06281357095350094,
|
| 538 |
-
"recall_mean": 0.5757575757575758,
|
| 539 |
-
"recall_sample_sd": 0.262431940540739,
|
| 540 |
-
"f1_mean": 0.35477582846003897,
|
| 541 |
-
"f1_sample_sd": 0.11479479036518875
|
| 542 |
-
},
|
| 543 |
-
{
|
| 544 |
-
"arm": "structural_copy_control",
|
| 545 |
-
"label": "Appeal_to_fear-prejudice",
|
| 546 |
-
"support": 24,
|
| 547 |
-
"precision_mean": 0.5104813315339631,
|
| 548 |
-
"precision_sample_sd": 0.038427508996734555,
|
| 549 |
-
"recall_mean": 0.8611111111111112,
|
| 550 |
-
"recall_sample_sd": 0.024056261216234387,
|
| 551 |
-
"f1_mean": 0.6401745363877202,
|
| 552 |
-
"f1_sample_sd": 0.029305467347150463
|
| 553 |
-
},
|
| 554 |
-
{
|
| 555 |
-
"arm": "structural_copy_control",
|
| 556 |
-
"label": "Bandwagon,Reductio_ad_hitlerum",
|
| 557 |
-
"support": 7,
|
| 558 |
-
"precision_mean": 0.2876623376623376,
|
| 559 |
-
"precision_sample_sd": 0.061386823077618426,
|
| 560 |
-
"recall_mean": 0.8571428571428571,
|
| 561 |
-
"recall_sample_sd": 0.14285714285714285,
|
| 562 |
-
"f1_mean": 0.4306391777656145,
|
| 563 |
-
"f1_sample_sd": 0.08686392624022733
|
| 564 |
-
},
|
| 565 |
-
{
|
| 566 |
-
"arm": "structural_copy_control",
|
| 567 |
-
"label": "Black-and-White_Fallacy",
|
| 568 |
-
"support": 12,
|
| 569 |
-
"precision_mean": 0.31339031339031337,
|
| 570 |
-
"precision_sample_sd": 0.08301311842925053,
|
| 571 |
-
"recall_mean": 0.5277777777777778,
|
| 572 |
-
"recall_sample_sd": 0.2678791878053599,
|
| 573 |
-
"f1_mean": 0.38778427550357375,
|
| 574 |
-
"f1_sample_sd": 0.13069755090515112
|
| 575 |
-
},
|
| 576 |
-
{
|
| 577 |
-
"arm": "structural_copy_control",
|
| 578 |
-
"label": "Causal_Oversimplification",
|
| 579 |
-
"support": 18,
|
| 580 |
-
"precision_mean": 0.4464201043148412,
|
| 581 |
-
"precision_sample_sd": 0.013538446216471536,
|
| 582 |
-
"recall_mean": 0.9259259259259259,
|
| 583 |
-
"recall_sample_sd": 0.03207501495497923,
|
| 584 |
-
"f1_mean": 0.6023809523809524,
|
| 585 |
-
"f1_sample_sd": 0.018643640071557464
|
| 586 |
-
},
|
| 587 |
-
{
|
| 588 |
-
"arm": "structural_copy_control",
|
| 589 |
-
"label": "Doubt",
|
| 590 |
-
"support": 22,
|
| 591 |
-
"precision_mean": 0.4455314009661836,
|
| 592 |
-
"precision_sample_sd": 0.04067072694857813,
|
| 593 |
-
"recall_mean": 1.0,
|
| 594 |
-
"recall_sample_sd": 0.0,
|
| 595 |
-
"f1_mean": 0.615686274509804,
|
| 596 |
-
"f1_sample_sd": 0.039427200694482675
|
| 597 |
-
},
|
| 598 |
-
{
|
| 599 |
-
"arm": "structural_copy_control",
|
| 600 |
-
"label": "Exaggeration,Minimisation",
|
| 601 |
-
"support": 22,
|
| 602 |
-
"precision_mean": 0.4364176150420584,
|
| 603 |
-
"precision_sample_sd": 0.025093958106097894,
|
| 604 |
-
"recall_mean": 0.8787878787878788,
|
| 605 |
-
"recall_sample_sd": 0.05248638810814775,
|
| 606 |
-
"f1_mean": 0.5829803047194351,
|
| 607 |
-
"f1_sample_sd": 0.030899288356345502
|
| 608 |
-
},
|
| 609 |
-
{
|
| 610 |
-
"arm": "structural_copy_control",
|
| 611 |
-
"label": "Flag-Waving",
|
| 612 |
-
"support": 17,
|
| 613 |
-
"precision_mean": 0.3814814814814815,
|
| 614 |
-
"precision_sample_sd": 0.05010277503136551,
|
| 615 |
-
"recall_mean": 0.803921568627451,
|
| 616 |
-
"recall_sample_sd": 0.17970885078258195,
|
| 617 |
-
"f1_mean": 0.5138595287119652,
|
| 618 |
-
"f1_sample_sd": 0.06400931989326974
|
| 619 |
-
},
|
| 620 |
-
{
|
| 621 |
-
"arm": "structural_copy_control",
|
| 622 |
-
"label": "Loaded_Language",
|
| 623 |
-
"support": 45,
|
| 624 |
-
"precision_mean": 0.8492581841638446,
|
| 625 |
-
"precision_sample_sd": 0.016026591848331686,
|
| 626 |
-
"recall_mean": 1.0,
|
| 627 |
-
"recall_sample_sd": 0.0,
|
| 628 |
-
"f1_mean": 0.9184311025253588,
|
| 629 |
-
"f1_sample_sd": 0.009372233868132662
|
| 630 |
-
},
|
| 631 |
-
{
|
| 632 |
-
"arm": "structural_copy_control",
|
| 633 |
-
"label": "Name_Calling,Labeling",
|
| 634 |
-
"support": 33,
|
| 635 |
-
"precision_mean": 0.6375,
|
| 636 |
-
"precision_sample_sd": 0.03410563654946854,
|
| 637 |
-
"recall_mean": 0.9696969696969697,
|
| 638 |
-
"recall_sample_sd": 0.030303030303030276,
|
| 639 |
-
"f1_mean": 0.7685185185185185,
|
| 640 |
-
"f1_sample_sd": 0.02023900779105559
|
| 641 |
-
},
|
| 642 |
-
{
|
| 643 |
-
"arm": "structural_copy_control",
|
| 644 |
-
"label": "Repetition",
|
| 645 |
-
"support": 23,
|
| 646 |
-
"precision_mean": 0.44863933452168747,
|
| 647 |
-
"precision_sample_sd": 0.01520014564765896,
|
| 648 |
-
"recall_mean": 0.9420289855072463,
|
| 649 |
-
"recall_sample_sd": 0.06641414050660639,
|
| 650 |
-
"f1_mean": 0.607248835423033,
|
| 651 |
-
"f1_sample_sd": 0.019858627551810427
|
| 652 |
-
},
|
| 653 |
-
{
|
| 654 |
-
"arm": "structural_copy_control",
|
| 655 |
-
"label": "Slogans",
|
| 656 |
-
"support": 14,
|
| 657 |
-
"precision_mean": 0.3274318274318274,
|
| 658 |
-
"precision_sample_sd": 0.05434598572773684,
|
| 659 |
-
"recall_mean": 0.4523809523809524,
|
| 660 |
-
"recall_sample_sd": 0.1486904285332952,
|
| 661 |
-
"f1_mean": 0.3715828924162258,
|
| 662 |
-
"f1_sample_sd": 0.07106660236343577
|
| 663 |
-
},
|
| 664 |
-
{
|
| 665 |
-
"arm": "structural_copy_control",
|
| 666 |
-
"label": "Thought-terminating_Cliches",
|
| 667 |
-
"support": 7,
|
| 668 |
-
"precision_mean": 0.1142156862745098,
|
| 669 |
-
"precision_sample_sd": 0.012848374923608973,
|
| 670 |
-
"recall_mean": 0.38095238095238093,
|
| 671 |
-
"recall_sample_sd": 0.1649572197684645,
|
| 672 |
-
"f1_mean": 0.17331433998100665,
|
| 673 |
-
"f1_sample_sd": 0.029065880233243847
|
| 674 |
-
},
|
| 675 |
-
{
|
| 676 |
-
"arm": "structural_copy_control",
|
| 677 |
-
"label": "Whataboutism,Straw_Men,Red_Herring",
|
| 678 |
-
"support": 10,
|
| 679 |
-
"precision_mean": 0.19344598054275472,
|
| 680 |
-
"precision_sample_sd": 0.03373872471827226,
|
| 681 |
-
"recall_mean": 0.5666666666666667,
|
| 682 |
-
"recall_sample_sd": 0.2081665999466133,
|
| 683 |
-
"f1_mean": 0.28584083690299267,
|
| 684 |
-
"f1_sample_sd": 0.06078854830133676
|
| 685 |
-
},
|
| 686 |
-
{
|
| 687 |
-
"arm": "existing_specialized_separate_lineage",
|
| 688 |
-
"label": "Appeal_to_Authority",
|
| 689 |
-
"support": 11,
|
| 690 |
-
"precision_mean": 0.37836021505376344,
|
| 691 |
-
"precision_sample_sd": 0.10546365636105155,
|
| 692 |
-
"recall_mean": 0.6666666666666666,
|
| 693 |
-
"recall_sample_sd": 0.2287828616748712,
|
| 694 |
-
"f1_mean": 0.46885361552028215,
|
| 695 |
-
"f1_sample_sd": 0.09502747629890008
|
| 696 |
-
},
|
| 697 |
-
{
|
| 698 |
-
"arm": "existing_specialized_separate_lineage",
|
| 699 |
-
"label": "Appeal_to_fear-prejudice",
|
| 700 |
-
"support": 24,
|
| 701 |
-
"precision_mean": 0.525708635464733,
|
| 702 |
-
"precision_sample_sd": 0.02235198659615819,
|
| 703 |
-
"recall_mean": 0.8888888888888888,
|
| 704 |
-
"recall_sample_sd": 0.048112522432468774,
|
| 705 |
-
"f1_mean": 0.6599065351235072,
|
| 706 |
-
"f1_sample_sd": 0.015362381542320084
|
| 707 |
-
},
|
| 708 |
-
{
|
| 709 |
-
"arm": "existing_specialized_separate_lineage",
|
| 710 |
-
"label": "Bandwagon,Reductio_ad_hitlerum",
|
| 711 |
-
"support": 7,
|
| 712 |
-
"precision_mean": 0.44485144485144484,
|
| 713 |
-
"precision_sample_sd": 0.016487284575115283,
|
| 714 |
-
"recall_mean": 0.7619047619047619,
|
| 715 |
-
"recall_sample_sd": 0.1649572197684645,
|
| 716 |
-
"f1_mean": 0.5571428571428572,
|
| 717 |
-
"f1_sample_sd": 0.051507875363771265
|
| 718 |
-
},
|
| 719 |
-
{
|
| 720 |
-
"arm": "existing_specialized_separate_lineage",
|
| 721 |
-
"label": "Black-and-White_Fallacy",
|
| 722 |
-
"support": 12,
|
| 723 |
-
"precision_mean": 0.4662644544997486,
|
| 724 |
-
"precision_sample_sd": 0.004538274146363688,
|
| 725 |
-
"recall_mean": 0.5833333333333334,
|
| 726 |
-
"recall_sample_sd": 0.08333333333333331,
|
| 727 |
-
"f1_mean": 0.516747552149851,
|
| 728 |
-
"f1_sample_sd": 0.03589484965697345
|
| 729 |
-
},
|
| 730 |
-
{
|
| 731 |
-
"arm": "existing_specialized_separate_lineage",
|
| 732 |
-
"label": "Causal_Oversimplification",
|
| 733 |
-
"support": 18,
|
| 734 |
-
"precision_mean": 0.4472934472934473,
|
| 735 |
-
"precision_sample_sd": 0.013055771210700413,
|
| 736 |
-
"recall_mean": 0.8333333333333333,
|
| 737 |
-
"recall_sample_sd": 0.1469861839480328,
|
| 738 |
-
"f1_mean": 0.5781794553724379,
|
| 739 |
-
"f1_sample_sd": 0.028407563015813122
|
| 740 |
-
},
|
| 741 |
-
{
|
| 742 |
-
"arm": "existing_specialized_separate_lineage",
|
| 743 |
-
"label": "Doubt",
|
| 744 |
-
"support": 22,
|
| 745 |
-
"precision_mean": 0.5045661001171619,
|
| 746 |
-
"precision_sample_sd": 0.023581174969843577,
|
| 747 |
-
"recall_mean": 1.0,
|
| 748 |
-
"recall_sample_sd": 0.0,
|
| 749 |
-
"f1_mean": 0.6704939668174963,
|
| 750 |
-
"f1_sample_sd": 0.020973132363659372
|
| 751 |
-
},
|
| 752 |
-
{
|
| 753 |
-
"arm": "existing_specialized_separate_lineage",
|
| 754 |
-
"label": "Exaggeration,Minimisation",
|
| 755 |
-
"support": 22,
|
| 756 |
-
"precision_mean": 0.44589793281653745,
|
| 757 |
-
"precision_sample_sd": 0.010987847993006481,
|
| 758 |
-
"recall_mean": 0.9393939393939394,
|
| 759 |
-
"recall_sample_sd": 0.06943296507508849,
|
| 760 |
-
"f1_mean": 0.6043956043956044,
|
| 761 |
-
"f1_sample_sd": 0.022305256187015846
|
| 762 |
-
},
|
| 763 |
-
{
|
| 764 |
-
"arm": "existing_specialized_separate_lineage",
|
| 765 |
-
"label": "Flag-Waving",
|
| 766 |
-
"support": 17,
|
| 767 |
-
"precision_mean": 0.5576388888888889,
|
| 768 |
-
"precision_sample_sd": 0.03705367103320583,
|
| 769 |
-
"recall_mean": 0.8823529411764706,
|
| 770 |
-
"recall_sample_sd": 0.11764705882352944,
|
| 771 |
-
"f1_mean": 0.680769868923179,
|
| 772 |
-
"f1_sample_sd": 0.04164658769406446
|
| 773 |
-
},
|
| 774 |
-
{
|
| 775 |
-
"arm": "existing_specialized_separate_lineage",
|
| 776 |
-
"label": "Loaded_Language",
|
| 777 |
-
"support": 45,
|
| 778 |
-
"precision_mean": 0.8387650085763294,
|
| 779 |
-
"precision_sample_sd": 0.01782556577257849,
|
| 780 |
-
"recall_mean": 1.0,
|
| 781 |
-
"recall_sample_sd": 0.0,
|
| 782 |
-
"f1_mean": 0.9122448979591837,
|
| 783 |
-
"f1_sample_sd": 0.01060439269940129
|
| 784 |
-
},
|
| 785 |
-
{
|
| 786 |
-
"arm": "existing_specialized_separate_lineage",
|
| 787 |
-
"label": "Name_Calling,Labeling",
|
| 788 |
-
"support": 33,
|
| 789 |
-
"precision_mean": 0.6957104190221673,
|
| 790 |
-
"precision_sample_sd": 0.019201490809269257,
|
| 791 |
-
"recall_mean": 0.9696969696969697,
|
| 792 |
-
"recall_sample_sd": 0.030303030303030276,
|
| 793 |
-
"f1_mean": 0.8101049442821594,
|
| 794 |
-
"f1_sample_sd": 0.022092733328802537
|
| 795 |
-
},
|
| 796 |
-
{
|
| 797 |
-
"arm": "existing_specialized_separate_lineage",
|
| 798 |
-
"label": "Repetition",
|
| 799 |
-
"support": 23,
|
| 800 |
-
"precision_mean": 0.4781557067271353,
|
| 801 |
-
"precision_sample_sd": 0.009897987322673213,
|
| 802 |
-
"recall_mean": 0.9420289855072463,
|
| 803 |
-
"recall_sample_sd": 0.06641414050660639,
|
| 804 |
-
"f1_mean": 0.6337774426009719,
|
| 805 |
-
"f1_sample_sd": 0.016444118333485254
|
| 806 |
-
},
|
| 807 |
-
{
|
| 808 |
-
"arm": "existing_specialized_separate_lineage",
|
| 809 |
-
"label": "Slogans",
|
| 810 |
-
"support": 14,
|
| 811 |
-
"precision_mean": 0.4334215167548501,
|
| 812 |
-
"precision_sample_sd": 0.05811581861282867,
|
| 813 |
-
"recall_mean": 0.6904761904761905,
|
| 814 |
-
"recall_sample_sd": 0.1649572197684645,
|
| 815 |
-
"f1_mean": 0.5201316298877274,
|
| 816 |
-
"f1_sample_sd": 0.018567913895656114
|
| 817 |
-
},
|
| 818 |
-
{
|
| 819 |
-
"arm": "existing_specialized_separate_lineage",
|
| 820 |
-
"label": "Thought-terminating_Cliches",
|
| 821 |
-
"support": 7,
|
| 822 |
-
"precision_mean": 0.18446115288220552,
|
| 823 |
-
"precision_sample_sd": 0.03641251252864963,
|
| 824 |
-
"recall_mean": 0.6666666666666666,
|
| 825 |
-
"recall_sample_sd": 0.1649572197684645,
|
| 826 |
-
"f1_mean": 0.2868626868626869,
|
| 827 |
-
"f1_sample_sd": 0.0511620864429468
|
| 828 |
-
},
|
| 829 |
-
{
|
| 830 |
-
"arm": "existing_specialized_separate_lineage",
|
| 831 |
-
"label": "Whataboutism,Straw_Men,Red_Herring",
|
| 832 |
-
"support": 10,
|
| 833 |
-
"precision_mean": 0.2777777777777778,
|
| 834 |
-
"precision_sample_sd": 0.0481125224324688,
|
| 835 |
-
"recall_mean": 0.4,
|
| 836 |
-
"recall_sample_sd": 0.0,
|
| 837 |
-
"f1_mean": 0.32634032634032634,
|
| 838 |
-
"f1_sample_sd": 0.03229931575886014
|
| 839 |
-
}
|
| 840 |
-
],
|
| 841 |
-
"compression_ladder": [
|
| 842 |
-
{
|
| 843 |
-
"name": "24L upstream text teacher",
|
| 844 |
-
"layers": 24,
|
| 845 |
-
"parameters": 752393024,
|
| 846 |
-
"weight_bytes": 1504825632,
|
| 847 |
-
"weight_mib": 1435.1135559082031,
|
| 848 |
-
"weight_sha256": "30783da4651259c26b8a204821ea426b9526431e0c14da0d1c6fa284e7414e5e",
|
| 849 |
-
"kd_stage": "",
|
| 850 |
-
"mean_total_loss": "",
|
| 851 |
-
"mean_interface_loss": "",
|
| 852 |
-
"mean_final_loss": ""
|
| 853 |
-
},
|
| 854 |
-
{
|
| 855 |
-
"name": "8L task-free KD",
|
| 856 |
-
"layers": 8,
|
| 857 |
-
"parameters": 420318144,
|
| 858 |
-
"weight_bytes": 840647880,
|
| 859 |
-
"weight_mib": 801.7042922973633,
|
| 860 |
-
"weight_sha256": "086fbaf9a4838ecde56cd3541c9c3212b22d0c5ee745cf2c1c5e0bb31e0ec22c",
|
| 861 |
-
"kd_stage": "24to8",
|
| 862 |
-
"mean_total_loss": 0.17622653172702485,
|
| 863 |
-
"mean_interface_loss": 0.21562811763578793,
|
| 864 |
-
"mean_final_loss": 0.16626427527853593
|
| 865 |
-
},
|
| 866 |
-
{
|
| 867 |
-
"name": "6L task-free KD",
|
| 868 |
-
"layers": 6,
|
| 869 |
-
"parameters": 377207424,
|
| 870 |
-
"weight_bytes": 754423424,
|
| 871 |
-
"weight_mib": 719.4742431640625,
|
| 872 |
-
"weight_sha256": "492749c20f77a42f6bc993f0927a82217b94f872b7eacf25612f7520c9bc28ce",
|
| 873 |
-
"kd_stage": "8to6",
|
| 874 |
-
"mean_total_loss": 0.03382104352249371,
|
| 875 |
-
"mean_interface_loss": 0.04486069045515251,
|
| 876 |
-
"mean_final_loss": 0.031147060785087888
|
| 877 |
-
},
|
| 878 |
-
{
|
| 879 |
-
"name": "4L task-free KD root",
|
| 880 |
-
"layers": 4,
|
| 881 |
-
"parameters": 334096704,
|
| 882 |
-
"weight_bytes": 668198976,
|
| 883 |
-
"weight_mib": 637.2442016601562,
|
| 884 |
-
"weight_sha256": "2732c616772fe320cdea228ab4554981418b1b2bf615c4183fb1ac8e6e2168d3",
|
| 885 |
-
"kd_stage": "6to4",
|
| 886 |
-
"mean_total_loss": 0.047020394468859195,
|
| 887 |
-
"mean_interface_loss": 0.07048421185527332,
|
| 888 |
-
"mean_final_loss": 0.043119939065377366
|
| 889 |
-
}
|
| 890 |
-
]
|
| 891 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@@ -1,7 +1,7 @@
|
|
| 1 |
{
|
| 2 |
"schema_version": "standalone4l-base-finalization-v1",
|
| 3 |
-
"source": "
|
| 4 |
-
"output": "
|
| 5 |
"source_hashes": {
|
| 6 |
"model.safetensors": "694ce83253652ec1287deaab7cc328ed3ff735ca25e1cef682dca82c80892aa0",
|
| 7 |
"config.json": "7e95a39a6efe3d34d8d79e27bcebc4068cf91f23b50f4e3bc9a828a0cbb38cf3",
|
|
|
|
| 1 |
{
|
| 2 |
"schema_version": "standalone4l-base-finalization-v1",
|
| 3 |
+
"source": "results/standalone4l_base/20260920-taskblind-grid/N32768/base-4l",
|
| 4 |
+
"output": "results/standalone4l_base/20260920-taskblind-grid/N32768/base-4l-canonical",
|
| 5 |
"source_hashes": {
|
| 6 |
"model.safetensors": "694ce83253652ec1287deaab7cc328ed3ff735ca25e1cef682dca82c80892aa0",
|
| 7 |
"config.json": "7e95a39a6efe3d34d8d79e27bcebc4068cf91f23b50f4e3bc9a828a0cbb38cf3",
|
|
@@ -1,7 +1,7 @@
|
|
| 1 |
{
|
| 2 |
"schema_version": "standalone4l-base-finalization-v1",
|
| 3 |
-
"source": "
|
| 4 |
-
"output": "
|
| 5 |
"source_hashes": {
|
| 6 |
"model.safetensors": "694ce83253652ec1287deaab7cc328ed3ff735ca25e1cef682dca82c80892aa0",
|
| 7 |
"config.json": "7e95a39a6efe3d34d8d79e27bcebc4068cf91f23b50f4e3bc9a828a0cbb38cf3",
|
|
|
|
| 1 |
{
|
| 2 |
"schema_version": "standalone4l-base-finalization-v1",
|
| 3 |
+
"source": "results/standalone4l_base/20260920-taskblind-grid/N32768/base-4l",
|
| 4 |
+
"output": "results/standalone4l_base/20260920-taskblind-grid/N32768/base-4l-canonical",
|
| 5 |
"source_hashes": {
|
| 6 |
"model.safetensors": "694ce83253652ec1287deaab7cc328ed3ff735ca25e1cef682dca82c80892aa0",
|
| 7 |
"config.json": "7e95a39a6efe3d34d8d79e27bcebc4068cf91f23b50f4e3bc9a828a0cbb38cf3",
|
|
@@ -3,11 +3,11 @@
|
|
| 3 |
"status": "completed",
|
| 4 |
"timestamp_kst": "2026-09-20T07:56:37.853994+09:00",
|
| 5 |
"stage": "6to4",
|
| 6 |
-
"teacher": "
|
| 7 |
-
"student_initial": "
|
| 8 |
-
"output": "
|
| 9 |
"data": {
|
| 10 |
-
"path": "
|
| 11 |
"sha256": "74da360f23826045b3e6ac6375411fdb15f003030aa74f2596ed08b857cb9212",
|
| 12 |
"rows_used": 4096,
|
| 13 |
"labels_read": false,
|
|
|
|
| 3 |
"status": "completed",
|
| 4 |
"timestamp_kst": "2026-09-20T07:56:37.853994+09:00",
|
| 5 |
"stage": "6to4",
|
| 6 |
+
"teacher": "results/standalone4l_base/20260920-taskblind-grid/N32768/base-6l",
|
| 7 |
+
"student_initial": "results/standalone4l_base/20260920-taskblind-grid/N32768/init-4l",
|
| 8 |
+
"output": "results/standalone4l_base/20260920-taskblind-grid/N32768/base-4l",
|
| 9 |
"data": {
|
| 10 |
+
"path": "data/public_corpora/wikitext-103-raw-v1-b08601e04326/train-00000-of-00002.parquet",
|
| 11 |
"sha256": "74da360f23826045b3e6ac6375411fdb15f003030aa74f2596ed08b857cb9212",
|
| 12 |
"rows_used": 4096,
|
| 13 |
"labels_read": false,
|
|
@@ -3,11 +3,11 @@
|
|
| 3 |
"status": "completed",
|
| 4 |
"timestamp_kst": "2026-09-20T07:45:30.667739+09:00",
|
| 5 |
"stage": "24to8",
|
| 6 |
-
"teacher": "
|
| 7 |
-
"student_initial": "
|
| 8 |
-
"output": "
|
| 9 |
"data": {
|
| 10 |
-
"path": "
|
| 11 |
"sha256": "74da360f23826045b3e6ac6375411fdb15f003030aa74f2596ed08b857cb9212",
|
| 12 |
"rows_used": 4096,
|
| 13 |
"labels_read": false,
|
|
|
|
| 3 |
"status": "completed",
|
| 4 |
"timestamp_kst": "2026-09-20T07:45:30.667739+09:00",
|
| 5 |
"stage": "24to8",
|
| 6 |
+
"teacher": "results/standalone4l_base/20260920-taskblind-grid/teacher-24l-N32768",
|
| 7 |
+
"student_initial": "results/standalone4l_base/20260920-taskblind-grid/N32768/init-8l",
|
| 8 |
+
"output": "results/standalone4l_base/20260920-taskblind-grid/N32768/base-8l",
|
| 9 |
"data": {
|
| 10 |
+
"path": "data/public_corpora/wikitext-103-raw-v1-b08601e04326/train-00000-of-00002.parquet",
|
| 11 |
"sha256": "74da360f23826045b3e6ac6375411fdb15f003030aa74f2596ed08b857cb9212",
|
| 12 |
"rows_used": 4096,
|
| 13 |
"labels_read": false,
|
|
@@ -3,11 +3,11 @@
|
|
| 3 |
"status": "completed",
|
| 4 |
"timestamp_kst": "2026-09-20T07:56:37.853994+09:00",
|
| 5 |
"stage": "6to4",
|
| 6 |
-
"teacher": "
|
| 7 |
-
"student_initial": "
|
| 8 |
-
"output": "
|
| 9 |
"data": {
|
| 10 |
-
"path": "
|
| 11 |
"sha256": "74da360f23826045b3e6ac6375411fdb15f003030aa74f2596ed08b857cb9212",
|
| 12 |
"rows_used": 4096,
|
| 13 |
"labels_read": false,
|
|
|
|
| 3 |
"status": "completed",
|
| 4 |
"timestamp_kst": "2026-09-20T07:56:37.853994+09:00",
|
| 5 |
"stage": "6to4",
|
| 6 |
+
"teacher": "results/standalone4l_base/20260920-taskblind-grid/N32768/base-6l",
|
| 7 |
+
"student_initial": "results/standalone4l_base/20260920-taskblind-grid/N32768/init-4l",
|
| 8 |
+
"output": "results/standalone4l_base/20260920-taskblind-grid/N32768/base-4l",
|
| 9 |
"data": {
|
| 10 |
+
"path": "data/public_corpora/wikitext-103-raw-v1-b08601e04326/train-00000-of-00002.parquet",
|
| 11 |
"sha256": "74da360f23826045b3e6ac6375411fdb15f003030aa74f2596ed08b857cb9212",
|
| 12 |
"rows_used": 4096,
|
| 13 |
"labels_read": false,
|
|
@@ -3,11 +3,11 @@
|
|
| 3 |
"status": "completed",
|
| 4 |
"timestamp_kst": "2026-09-20T07:52:32.717463+09:00",
|
| 5 |
"stage": "8to6",
|
| 6 |
-
"teacher": "
|
| 7 |
-
"student_initial": "
|
| 8 |
-
"output": "
|
| 9 |
"data": {
|
| 10 |
-
"path": "
|
| 11 |
"sha256": "74da360f23826045b3e6ac6375411fdb15f003030aa74f2596ed08b857cb9212",
|
| 12 |
"rows_used": 4096,
|
| 13 |
"labels_read": false,
|
|
|
|
| 3 |
"status": "completed",
|
| 4 |
"timestamp_kst": "2026-09-20T07:52:32.717463+09:00",
|
| 5 |
"stage": "8to6",
|
| 6 |
+
"teacher": "results/standalone4l_base/20260920-taskblind-grid/N32768/base-8l",
|
| 7 |
+
"student_initial": "results/standalone4l_base/20260920-taskblind-grid/N32768/init-6l",
|
| 8 |
+
"output": "results/standalone4l_base/20260920-taskblind-grid/N32768/base-6l",
|
| 9 |
"data": {
|
| 10 |
+
"path": "data/public_corpora/wikitext-103-raw-v1-b08601e04326/train-00000-of-00002.parquet",
|
| 11 |
"sha256": "74da360f23826045b3e6ac6375411fdb15f003030aa74f2596ed08b857cb9212",
|
| 12 |
"rows_used": 4096,
|
| 13 |
"labels_read": false,
|
|
@@ -1,675 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"schema": "qwen35-unified-classification-base-release-v3-v128k-update",
|
| 3 |
-
"status": "complete",
|
| 4 |
-
"repo_id": "mp-juuuns/qwen35-standalone4l-classification-base",
|
| 5 |
-
"old_repo_preserved": {
|
| 6 |
-
"repo_id": "mp-juuuns/qwen35-standalone4l-propaganda-classifier",
|
| 7 |
-
"revision": "e7e0ff16828052687d2e8dd7849e7a521629cf38"
|
| 8 |
-
},
|
| 9 |
-
"root_role": "task_agnostic_headless_classification_backbone",
|
| 10 |
-
"root_weight_sha256": "7aa658a1adef9665fa86c0cdd97d0e32c57e3dd6be618eaaf7794448930d387c",
|
| 11 |
-
"root_multilingual_metadata_scope": "inherited tokenizer/architecture coverage; no direct multilingual downstream evaluation",
|
| 12 |
-
"benchmark_schema": "standalone4l-complete-transfer-benchmark-v2",
|
| 13 |
-
"figure_schema": "standalone4l-matplotlib-figures-v1",
|
| 14 |
-
"nested_specialized_is_separate_lineage": true,
|
| 15 |
-
"source_data_redistributed": false,
|
| 16 |
-
"files_before_manifest": [
|
| 17 |
-
{
|
| 18 |
-
"path": ".gitattributes",
|
| 19 |
-
"role": "Hugging Face LFS policy",
|
| 20 |
-
"bytes": 50,
|
| 21 |
-
"sha256": "ee05350c7fda0af4da116020dd7b976d731de3beed4f5e717d5bcc95ead24550",
|
| 22 |
-
"source": "generated"
|
| 23 |
-
},
|
| 24 |
-
{
|
| 25 |
-
"path": "LICENSE",
|
| 26 |
-
"role": "Apache-2.0 license",
|
| 27 |
-
"bytes": 11418,
|
| 28 |
-
"sha256": "77fd4710def9ec3c0f6225800e0235f15a425abd4a8b03559127fcd782612049",
|
| 29 |
-
"source": "references/software/transformers-5.13.0/LICENSE"
|
| 30 |
-
},
|
| 31 |
-
{
|
| 32 |
-
"path": "README.md",
|
| 33 |
-
"role": "root model card",
|
| 34 |
-
"bytes": 16305,
|
| 35 |
-
"sha256": "e8456e330ad315f08a6c37e308b6e3bbe54b31f6d65fc37a6a1625386d99a529",
|
| 36 |
-
"source": "docs/huggingface/UNIFIED_BASE_RELEASE_MODEL_CARD.md"
|
| 37 |
-
},
|
| 38 |
-
{
|
| 39 |
-
"path": "benchmark/BENCHMARK_CARD.md",
|
| 40 |
-
"role": "generated benchmark report",
|
| 41 |
-
"bytes": 3173,
|
| 42 |
-
"sha256": "bf4073de38c3905effd2dcce6f38894f325c8a0a79bedfb2779f0545c86b9c9f",
|
| 43 |
-
"source": "generated"
|
| 44 |
-
},
|
| 45 |
-
{
|
| 46 |
-
"path": "benchmark/arm_summary.csv",
|
| 47 |
-
"role": "three-seed metric summary",
|
| 48 |
-
"bytes": 1655,
|
| 49 |
-
"sha256": "e01bd7ac59a3212e44b77e5fcfbc7c41ec91d74c998306d2438209d6ef1a3aa7",
|
| 50 |
-
"source": "generated"
|
| 51 |
-
},
|
| 52 |
-
{
|
| 53 |
-
"path": "benchmark/compression_ladder.csv",
|
| 54 |
-
"role": "24L to 4L compression evidence",
|
| 55 |
-
"bytes": 827,
|
| 56 |
-
"sha256": "2c544c3907c4f592d724f658850b8c6b9634845cb5b50a66ebd2ea6c5d48ce0b",
|
| 57 |
-
"source": "generated"
|
| 58 |
-
},
|
| 59 |
-
{
|
| 60 |
-
"path": "benchmark/figures/01_overall_transfer_quality.csv",
|
| 61 |
-
"role": "Matplotlib benchmark figures and source data",
|
| 62 |
-
"bytes": 500,
|
| 63 |
-
"sha256": "769a516e53417e16b61cf30d95a02f6c812c0b1536242cdbf836da779753efbd",
|
| 64 |
-
"source": "results/standalone4l_base/20260731-v1/huggingface_figures_v1/01_overall_transfer_quality.csv"
|
| 65 |
-
},
|
| 66 |
-
{
|
| 67 |
-
"path": "benchmark/figures/01_overall_transfer_quality.pdf",
|
| 68 |
-
"role": "Matplotlib benchmark figures and source data",
|
| 69 |
-
"bytes": 18071,
|
| 70 |
-
"sha256": "7221e0d2feb356e4d7f44aa0381e576254f734837e3f54fb0dfb98795ec19ded",
|
| 71 |
-
"source": "results/standalone4l_base/20260731-v1/huggingface_figures_v1/01_overall_transfer_quality.pdf"
|
| 72 |
-
},
|
| 73 |
-
{
|
| 74 |
-
"path": "benchmark/figures/01_overall_transfer_quality.png",
|
| 75 |
-
"role": "Matplotlib benchmark figures and source data",
|
| 76 |
-
"bytes": 82127,
|
| 77 |
-
"sha256": "6d35b5964791cfff2701e43831996a9aad7d607305ab3031b0782bed2ffffa66",
|
| 78 |
-
"source": "results/standalone4l_base/20260731-v1/huggingface_figures_v1/01_overall_transfer_quality.png"
|
| 79 |
-
},
|
| 80 |
-
{
|
| 81 |
-
"path": "benchmark/figures/02_per_label_transfer_metrics.csv",
|
| 82 |
-
"role": "Matplotlib benchmark figures and source data",
|
| 83 |
-
"bytes": 1169,
|
| 84 |
-
"sha256": "50a3ceb0e3146ba8cade9c86c4b59c7db6cbedfc41fac39a811d7f39f4bdeb0d",
|
| 85 |
-
"source": "results/standalone4l_base/20260731-v1/huggingface_figures_v1/02_per_label_transfer_metrics.csv"
|
| 86 |
-
},
|
| 87 |
-
{
|
| 88 |
-
"path": "benchmark/figures/02_per_label_transfer_metrics.pdf",
|
| 89 |
-
"role": "Matplotlib benchmark figures and source data",
|
| 90 |
-
"bytes": 21379,
|
| 91 |
-
"sha256": "10756816e8ad800f5d872907986680abf637e76fe47f9790f27645f8f231ae4a",
|
| 92 |
-
"source": "results/standalone4l_base/20260731-v1/huggingface_figures_v1/02_per_label_transfer_metrics.pdf"
|
| 93 |
-
},
|
| 94 |
-
{
|
| 95 |
-
"path": "benchmark/figures/02_per_label_transfer_metrics.png",
|
| 96 |
-
"role": "Matplotlib benchmark figures and source data",
|
| 97 |
-
"bytes": 138298,
|
| 98 |
-
"sha256": "5b558738b405a7b04bb30f38b6e53af21b2392ea12cad0d777e9d6d6985d3d04",
|
| 99 |
-
"source": "results/standalone4l_base/20260731-v1/huggingface_figures_v1/02_per_label_transfer_metrics.png"
|
| 100 |
-
},
|
| 101 |
-
{
|
| 102 |
-
"path": "benchmark/figures/03_quality_efficiency.csv",
|
| 103 |
-
"role": "Matplotlib benchmark figures and source data",
|
| 104 |
-
"bytes": 448,
|
| 105 |
-
"sha256": "f2a34cbcd700b2663cf761a5fb1dbd3aa99f310a96d4b4477a3a4529bee1a1b3",
|
| 106 |
-
"source": "results/standalone4l_base/20260731-v1/huggingface_figures_v1/03_quality_efficiency.csv"
|
| 107 |
-
},
|
| 108 |
-
{
|
| 109 |
-
"path": "benchmark/figures/03_quality_efficiency.pdf",
|
| 110 |
-
"role": "Matplotlib benchmark figures and source data",
|
| 111 |
-
"bytes": 21933,
|
| 112 |
-
"sha256": "18c81ce50fc7871b000c9e387f0f7d7780002cb2bd249153aadfa55a186342c0",
|
| 113 |
-
"source": "results/standalone4l_base/20260731-v1/huggingface_figures_v1/03_quality_efficiency.pdf"
|
| 114 |
-
},
|
| 115 |
-
{
|
| 116 |
-
"path": "benchmark/figures/03_quality_efficiency.png",
|
| 117 |
-
"role": "Matplotlib benchmark figures and source data",
|
| 118 |
-
"bytes": 99374,
|
| 119 |
-
"sha256": "08ed7ca8193d5f9065645fb22491ab88ee260d43566e9013aa0f98d7ab4dde87",
|
| 120 |
-
"source": "results/standalone4l_base/20260731-v1/huggingface_figures_v1/03_quality_efficiency.png"
|
| 121 |
-
},
|
| 122 |
-
{
|
| 123 |
-
"path": "benchmark/figures/04_compression_ladder.csv",
|
| 124 |
-
"role": "Matplotlib benchmark figures and source data",
|
| 125 |
-
"bytes": 352,
|
| 126 |
-
"sha256": "2dfac2b8330724ce46a22d8f3ddfeffc767f52b93a93cb1a833c89c5820ffccc",
|
| 127 |
-
"source": "results/standalone4l_base/20260731-v1/huggingface_figures_v1/04_compression_ladder.csv"
|
| 128 |
-
},
|
| 129 |
-
{
|
| 130 |
-
"path": "benchmark/figures/04_compression_ladder.pdf",
|
| 131 |
-
"role": "Matplotlib benchmark figures and source data",
|
| 132 |
-
"bytes": 19490,
|
| 133 |
-
"sha256": "953d6e356498127b2a607530fac2e41b07b4bf38ed0af945f778b784dc6f7b23",
|
| 134 |
-
"source": "results/standalone4l_base/20260731-v1/huggingface_figures_v1/04_compression_ladder.pdf"
|
| 135 |
-
},
|
| 136 |
-
{
|
| 137 |
-
"path": "benchmark/figures/04_compression_ladder.png",
|
| 138 |
-
"role": "Matplotlib benchmark figures and source data",
|
| 139 |
-
"bytes": 120809,
|
| 140 |
-
"sha256": "9f6c1d6fdefd45711fcfa57a7d0d91f6baaa3a5ee0b44a36554a9c74febe7090",
|
| 141 |
-
"source": "results/standalone4l_base/20260731-v1/huggingface_figures_v1/04_compression_ladder.png"
|
| 142 |
-
},
|
| 143 |
-
{
|
| 144 |
-
"path": "benchmark/figures/README.md",
|
| 145 |
-
"role": "Matplotlib benchmark figures and source data",
|
| 146 |
-
"bytes": 683,
|
| 147 |
-
"sha256": "7e00107e6576da117689dca0c164c497ce32e9d93c034331e2e161625afc2092",
|
| 148 |
-
"source": "results/standalone4l_base/20260731-v1/huggingface_figures_v1/README.md"
|
| 149 |
-
},
|
| 150 |
-
{
|
| 151 |
-
"path": "benchmark/figures/manifest.json",
|
| 152 |
-
"role": "Matplotlib benchmark figures and source data",
|
| 153 |
-
"bytes": 640,
|
| 154 |
-
"sha256": "41fbd7c746eb52cbb3ae7e7e5427ac1e509de4be65dd441548754e090f6bc74f",
|
| 155 |
-
"source": "results/standalone4l_base/20260731-v1/huggingface_figures_v1/manifest.json"
|
| 156 |
-
},
|
| 157 |
-
{
|
| 158 |
-
"path": "benchmark/figures/source_summary.json",
|
| 159 |
-
"role": "Matplotlib benchmark figures and source data",
|
| 160 |
-
"bytes": 26913,
|
| 161 |
-
"sha256": "3d2c9fcf9c6cc7da1a885968b70a42eaa38659d317c9d3290a2b3066f3f9ee6a",
|
| 162 |
-
"source": "results/standalone4l_base/20260731-v1/huggingface_figures_v1/source_summary.json"
|
| 163 |
-
},
|
| 164 |
-
{
|
| 165 |
-
"path": "benchmark/figures/v128k_compression.csv",
|
| 166 |
-
"role": "generated v128k update artifact",
|
| 167 |
-
"source": "generated",
|
| 168 |
-
"bytes": 198,
|
| 169 |
-
"sha256": "078b4f34897e0a1b9bcff66ce21ba0feb28d72c86eaa8cef608f110c48f52f2e"
|
| 170 |
-
},
|
| 171 |
-
{
|
| 172 |
-
"path": "benchmark/figures/v128k_quality_efficiency.csv",
|
| 173 |
-
"role": "generated v128k update artifact",
|
| 174 |
-
"source": "generated",
|
| 175 |
-
"bytes": 370,
|
| 176 |
-
"sha256": "014da12a435e3404b31470e2642d180dc111eb608fb8513fd993097019698b7a"
|
| 177 |
-
},
|
| 178 |
-
{
|
| 179 |
-
"path": "benchmark/figures/v128k_quality_storage.png",
|
| 180 |
-
"role": "generated v128k update artifact",
|
| 181 |
-
"source": "generated",
|
| 182 |
-
"bytes": 73388,
|
| 183 |
-
"sha256": "644e5b7469c0b4e0b63d5ecaa1c0f2fc8e07af8bb5d9e4d3ae5a88166de374f3"
|
| 184 |
-
},
|
| 185 |
-
{
|
| 186 |
-
"path": "benchmark/figures/v128k_vocabulary_size.png",
|
| 187 |
-
"role": "generated v128k update artifact",
|
| 188 |
-
"source": "generated",
|
| 189 |
-
"bytes": 52750,
|
| 190 |
-
"sha256": "66d5da977653e326893da3d34b5babb2e078ca1381753112e74ef76bace80193"
|
| 191 |
-
},
|
| 192 |
-
{
|
| 193 |
-
"path": "benchmark/per_label_metrics.csv",
|
| 194 |
-
"role": "per-label metrics by arm and seed",
|
| 195 |
-
"bytes": 12963,
|
| 196 |
-
"sha256": "ffc21be32b2106c7bf1df3854097a975ac370da01ac1b732eb5c2ba9075ddeda",
|
| 197 |
-
"source": "generated"
|
| 198 |
-
},
|
| 199 |
-
{
|
| 200 |
-
"path": "benchmark/per_label_summary.csv",
|
| 201 |
-
"role": "per-label three-seed summary",
|
| 202 |
-
"bytes": 7009,
|
| 203 |
-
"sha256": "c8e3876abf6cfe960c9332d3bc9d1fbde2f3213f3111351acf83b21bbb184049",
|
| 204 |
-
"source": "generated"
|
| 205 |
-
},
|
| 206 |
-
{
|
| 207 |
-
"path": "benchmark/reports/existing_specialized_separate_lineage-seed41.json",
|
| 208 |
-
"role": "normalized frozen benchmark report",
|
| 209 |
-
"bytes": 115793,
|
| 210 |
-
"sha256": "c1bd379680abcf50b47b7d23e4aafbc8b5e7f40a0b0e9700bde21a8ce49af803",
|
| 211 |
-
"source": "generated"
|
| 212 |
-
},
|
| 213 |
-
{
|
| 214 |
-
"path": "benchmark/reports/existing_specialized_separate_lineage-seed42.json",
|
| 215 |
-
"role": "normalized frozen benchmark report",
|
| 216 |
-
"bytes": 115780,
|
| 217 |
-
"sha256": "a30e38433d008db4aa7e5fbfd477ab80a0daf50e970f8e7b04397efc1a47843a",
|
| 218 |
-
"source": "generated"
|
| 219 |
-
},
|
| 220 |
-
{
|
| 221 |
-
"path": "benchmark/reports/existing_specialized_separate_lineage-seed43.json",
|
| 222 |
-
"role": "normalized frozen benchmark report",
|
| 223 |
-
"bytes": 115985,
|
| 224 |
-
"sha256": "96466290d09d27b8bc044a9f9a95765ce12b4d7b02dfb0123e436e7535623353",
|
| 225 |
-
"source": "generated"
|
| 226 |
-
},
|
| 227 |
-
{
|
| 228 |
-
"path": "benchmark/reports/structural_copy_control-seed41.json",
|
| 229 |
-
"role": "normalized frozen benchmark report",
|
| 230 |
-
"bytes": 115612,
|
| 231 |
-
"sha256": "816c3249d3ec281ef73d34dc3ab7576a48415d250c32977eb94ca37f73b0b810",
|
| 232 |
-
"source": "generated"
|
| 233 |
-
},
|
| 234 |
-
{
|
| 235 |
-
"path": "benchmark/reports/structural_copy_control-seed42.json",
|
| 236 |
-
"role": "normalized frozen benchmark report",
|
| 237 |
-
"bytes": 115225,
|
| 238 |
-
"sha256": "0b19eabe1e31d0012c754db3b31173f8a6fcdee30973f1a1bebd3a370d11064c",
|
| 239 |
-
"source": "generated"
|
| 240 |
-
},
|
| 241 |
-
{
|
| 242 |
-
"path": "benchmark/reports/structural_copy_control-seed43.json",
|
| 243 |
-
"role": "normalized frozen benchmark report",
|
| 244 |
-
"bytes": 115744,
|
| 245 |
-
"sha256": "6dde34173462eac50141ff082c18194f3233f8c06853532d9b12756562e69e2e",
|
| 246 |
-
"source": "generated"
|
| 247 |
-
},
|
| 248 |
-
{
|
| 249 |
-
"path": "benchmark/reports/task_agnostic_base-seed41.json",
|
| 250 |
-
"role": "normalized frozen benchmark report",
|
| 251 |
-
"bytes": 115885,
|
| 252 |
-
"sha256": "6b3a2e0372a83e1e71c2eca9596efc8cd1a3e39c28e25e3373f195431b5858a4",
|
| 253 |
-
"source": "generated"
|
| 254 |
-
},
|
| 255 |
-
{
|
| 256 |
-
"path": "benchmark/reports/task_agnostic_base-seed42.json",
|
| 257 |
-
"role": "normalized frozen benchmark report",
|
| 258 |
-
"bytes": 115674,
|
| 259 |
-
"sha256": "42dc788983716be0e9847aad87ff79c7e7297884f2635a8d098639be77326639",
|
| 260 |
-
"source": "generated"
|
| 261 |
-
},
|
| 262 |
-
{
|
| 263 |
-
"path": "benchmark/reports/task_agnostic_base-seed43.json",
|
| 264 |
-
"role": "normalized frozen benchmark report",
|
| 265 |
-
"bytes": 115062,
|
| 266 |
-
"sha256": "7ee52dc94a152a2f767dcff171f3c9cc1072fabd3ebc7b14db53617b7b1d33d5",
|
| 267 |
-
"source": "generated"
|
| 268 |
-
},
|
| 269 |
-
{
|
| 270 |
-
"path": "benchmark/resource_metrics.csv",
|
| 271 |
-
"role": "resource measurements by arm and seed",
|
| 272 |
-
"bytes": 1649,
|
| 273 |
-
"sha256": "bf5cefdec3b2f3646114bd18bbf8e5e39721bde768ba968d73f22e96fc9ca8cf",
|
| 274 |
-
"source": "generated"
|
| 275 |
-
},
|
| 276 |
-
{
|
| 277 |
-
"path": "benchmark/seed_metrics.csv",
|
| 278 |
-
"role": "overall metrics by arm and seed",
|
| 279 |
-
"bytes": 1175,
|
| 280 |
-
"sha256": "e0b08a958d25243158a03dff76ca153d2ded76cc3f1771e4046a0068688ad6a8",
|
| 281 |
-
"source": "generated"
|
| 282 |
-
},
|
| 283 |
-
{
|
| 284 |
-
"path": "benchmark/semeval-transfer-summary.json",
|
| 285 |
-
"role": "compatibility alias for complete transfer summary",
|
| 286 |
-
"bytes": 26913,
|
| 287 |
-
"sha256": "3d2c9fcf9c6cc7da1a885968b70a42eaa38659d317c9d3290a2b3066f3f9ee6a",
|
| 288 |
-
"source": "generated"
|
| 289 |
-
},
|
| 290 |
-
{
|
| 291 |
-
"path": "benchmark/summary.json",
|
| 292 |
-
"role": "complete machine-readable transfer benchmark",
|
| 293 |
-
"bytes": 26913,
|
| 294 |
-
"sha256": "3d2c9fcf9c6cc7da1a885968b70a42eaa38659d317c9d3290a2b3066f3f9ee6a",
|
| 295 |
-
"source": "generated"
|
| 296 |
-
},
|
| 297 |
-
{
|
| 298 |
-
"path": "config.json",
|
| 299 |
-
"role": "task-agnostic root checkpoint",
|
| 300 |
-
"bytes": 2156,
|
| 301 |
-
"sha256": "32967cd99c1386f0aecf0822e0e3f67164a72058e709d989b888928b0bbe46c5",
|
| 302 |
-
"source": "results/standalone4l_base/20260731-v1/base-4l-canonical/config.json"
|
| 303 |
-
},
|
| 304 |
-
{
|
| 305 |
-
"path": "distillation/README.md",
|
| 306 |
-
"role": "reusable public platform",
|
| 307 |
-
"bytes": 3432,
|
| 308 |
-
"sha256": "f5b7248f7781412e8b5d8f2fb3b0c309568e4e75baf55029b26a6fa985d9d608",
|
| 309 |
-
"source": "distillation/README.md"
|
| 310 |
-
},
|
| 311 |
-
{
|
| 312 |
-
"path": "distillation/configs/task_agnostic_base.yaml",
|
| 313 |
-
"role": "reusable public platform",
|
| 314 |
-
"bytes": 357,
|
| 315 |
-
"sha256": "adf8b6e3f0f5a5dfe042dcc09a7a8e7ef8cd1146462b089c5f6c9393e9a3131f",
|
| 316 |
-
"source": "distillation/configs/task_agnostic_base.yaml"
|
| 317 |
-
},
|
| 318 |
-
{
|
| 319 |
-
"path": "distillation/examples/labels.json",
|
| 320 |
-
"role": "reusable public platform",
|
| 321 |
-
"bytes": 36,
|
| 322 |
-
"sha256": "8b528ab38eb2e9f53f9ec2c50d4a0c860d87594067dc9b064b4bb47ef51f8f50",
|
| 323 |
-
"source": "distillation/examples/labels.json"
|
| 324 |
-
},
|
| 325 |
-
{
|
| 326 |
-
"path": "distillation/examples/multilabel_train.jsonl",
|
| 327 |
-
"role": "reusable public platform",
|
| 328 |
-
"bytes": 186,
|
| 329 |
-
"sha256": "daf5859bb93c474d766ecb0681a2585a6a8ed3cb5efcdd476c845727a0c3583c",
|
| 330 |
-
"source": "distillation/examples/multilabel_train.jsonl"
|
| 331 |
-
},
|
| 332 |
-
{
|
| 333 |
-
"path": "distillation/examples/single_label_train.jsonl",
|
| 334 |
-
"role": "reusable public platform",
|
| 335 |
-
"bytes": 171,
|
| 336 |
-
"sha256": "918490e8db364c6e601cfdcb09e28c272eafc8359ffe190645a4cbecdf06aed3",
|
| 337 |
-
"source": "distillation/examples/single_label_train.jsonl"
|
| 338 |
-
},
|
| 339 |
-
{
|
| 340 |
-
"path": "distillation/examples/unlabeled.jsonl",
|
| 341 |
-
"role": "reusable public platform",
|
| 342 |
-
"bytes": 183,
|
| 343 |
-
"sha256": "7a1db673674a003ef749b1fa90a99eff6db70c3189411b9cee574cc23d084860",
|
| 344 |
-
"source": "distillation/examples/unlabeled.jsonl"
|
| 345 |
-
},
|
| 346 |
-
{
|
| 347 |
-
"path": "distillation/pyproject.toml",
|
| 348 |
-
"role": "reusable public platform",
|
| 349 |
-
"bytes": 487,
|
| 350 |
-
"sha256": "3c5a0f9b48849381307a6b3c258729d4079bf5d42bb6ab5d28edbdda4a1a3592",
|
| 351 |
-
"source": "distillation/pyproject.toml"
|
| 352 |
-
},
|
| 353 |
-
{
|
| 354 |
-
"path": "distillation/qwen35_distill/__init__.py",
|
| 355 |
-
"role": "reusable public platform",
|
| 356 |
-
"bytes": 175,
|
| 357 |
-
"sha256": "2965e662310bf4197d1cbd4f9b3fcf3f383950dbfc4f033729a286d026709f3a",
|
| 358 |
-
"source": "distillation/qwen35_distill/__init__.py"
|
| 359 |
-
},
|
| 360 |
-
{
|
| 361 |
-
"path": "distillation/qwen35_distill/checkpoint.py",
|
| 362 |
-
"role": "reusable public platform",
|
| 363 |
-
"bytes": 12541,
|
| 364 |
-
"sha256": "d67bf7ed3bfb6cb622d6d35a99a4292f1191395270264e81a7d19283d206e5dc",
|
| 365 |
-
"source": "distillation/qwen35_distill/checkpoint.py"
|
| 366 |
-
},
|
| 367 |
-
{
|
| 368 |
-
"path": "distillation/qwen35_distill/classification.py",
|
| 369 |
-
"role": "reusable public platform",
|
| 370 |
-
"bytes": 14602,
|
| 371 |
-
"sha256": "ea06bec682622addc426def4f7e315b3d12f7f595af44f714cc811588cf64782",
|
| 372 |
-
"source": "distillation/qwen35_distill/classification.py"
|
| 373 |
-
},
|
| 374 |
-
{
|
| 375 |
-
"path": "distillation/qwen35_distill/cli.py",
|
| 376 |
-
"role": "reusable public platform",
|
| 377 |
-
"bytes": 8015,
|
| 378 |
-
"sha256": "d8fb1584484596f9c35db9c2c296dab08b2dd9bb1b7910a115710feee7562771",
|
| 379 |
-
"source": "distillation/qwen35_distill/cli.py"
|
| 380 |
-
},
|
| 381 |
-
{
|
| 382 |
-
"path": "distillation/qwen35_distill/layer_maps.py",
|
| 383 |
-
"role": "reusable public platform",
|
| 384 |
-
"bytes": 3103,
|
| 385 |
-
"sha256": "a9c49af0ee14b307a08c096bcd1ae55fa9bc67bdb807fa9383fc313246e7edd7",
|
| 386 |
-
"source": "distillation/qwen35_distill/layer_maps.py"
|
| 387 |
-
},
|
| 388 |
-
{
|
| 389 |
-
"path": "distillation/qwen35_distill/losses.py",
|
| 390 |
-
"role": "reusable public platform",
|
| 391 |
-
"bytes": 3794,
|
| 392 |
-
"sha256": "d9b135cf5b0adf4bbfc321200395fbd6112e49fd1a027b30262bcff1c4c0b654",
|
| 393 |
-
"source": "distillation/qwen35_distill/losses.py"
|
| 394 |
-
},
|
| 395 |
-
{
|
| 396 |
-
"path": "distillation/qwen35_distill/schema.py",
|
| 397 |
-
"role": "reusable public platform",
|
| 398 |
-
"bytes": 5607,
|
| 399 |
-
"sha256": "00731ae7838317105f9fff960e3da210b2a55954feb37b3ac793bcfc81fdab2d",
|
| 400 |
-
"source": "distillation/qwen35_distill/schema.py"
|
| 401 |
-
},
|
| 402 |
-
{
|
| 403 |
-
"path": "distillation/qwen35_distill/training.py",
|
| 404 |
-
"role": "reusable public platform",
|
| 405 |
-
"bytes": 10474,
|
| 406 |
-
"sha256": "4dc432acb70e03848d2e5964e8550065490fda3857aac32fe3b79895ef72026f",
|
| 407 |
-
"source": "distillation/qwen35_distill/training.py"
|
| 408 |
-
},
|
| 409 |
-
{
|
| 410 |
-
"path": "docs/RELEASE_CONTRACT.md",
|
| 411 |
-
"role": "release contract",
|
| 412 |
-
"bytes": 5381,
|
| 413 |
-
"sha256": "1b6e475b9d7d9af52bb8845de020303b1c6e89646d8e4f9be17dbe4ab2910417",
|
| 414 |
-
"source": "docs/huggingface/UNIFIED_BASE_RELEASE_CONTRACT.md"
|
| 415 |
-
},
|
| 416 |
-
{
|
| 417 |
-
"path": "model.safetensors",
|
| 418 |
-
"role": "task-agnostic root checkpoint",
|
| 419 |
-
"bytes": 421783648,
|
| 420 |
-
"sha256": "7aa658a1adef9665fa86c0cdd97d0e32c57e3dd6be618eaaf7794448930d387c",
|
| 421 |
-
"source": "results/standalone4l_base/20260731-v1/base-4l-canonical/model.safetensors"
|
| 422 |
-
},
|
| 423 |
-
{
|
| 424 |
-
"path": "models/semeval-propaganda/LICENSE",
|
| 425 |
-
"role": "byte-identical existing specialized release",
|
| 426 |
-
"bytes": 11418,
|
| 427 |
-
"sha256": "77fd4710def9ec3c0f6225800e0235f15a425abd4a8b03559127fcd782612049",
|
| 428 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/LICENSE"
|
| 429 |
-
},
|
| 430 |
-
{
|
| 431 |
-
"path": "models/semeval-propaganda/README.md",
|
| 432 |
-
"role": "specialized lineage card",
|
| 433 |
-
"bytes": 1403,
|
| 434 |
-
"sha256": "bf4ba7cc2514502a28f424507c41c1cfce5089c051ba7500cf46c1d13860722f",
|
| 435 |
-
"source": "generated"
|
| 436 |
-
},
|
| 437 |
-
{
|
| 438 |
-
"path": "models/semeval-propaganda/benchmark/final4l_aggregate.json",
|
| 439 |
-
"role": "existing specialized aggregate",
|
| 440 |
-
"bytes": 10882,
|
| 441 |
-
"sha256": "0bae347bb54f8089a5fa5450f8366b5c56eb3e598beaf9a87f4385363a402bcb",
|
| 442 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/benchmark/final4l_aggregate.json"
|
| 443 |
-
},
|
| 444 |
-
{
|
| 445 |
-
"path": "models/semeval-propaganda/benchmark/seed41.json",
|
| 446 |
-
"role": "existing specialized seed report",
|
| 447 |
-
"bytes": 115793,
|
| 448 |
-
"sha256": "c1bd379680abcf50b47b7d23e4aafbc8b5e7f40a0b0e9700bde21a8ce49af803",
|
| 449 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/benchmark/seed_reports/seed41.json"
|
| 450 |
-
},
|
| 451 |
-
{
|
| 452 |
-
"path": "models/semeval-propaganda/benchmark/seed42.json",
|
| 453 |
-
"role": "existing specialized seed report",
|
| 454 |
-
"bytes": 115780,
|
| 455 |
-
"sha256": "a30e38433d008db4aa7e5fbfd477ab80a0daf50e970f8e7b04397efc1a47843a",
|
| 456 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/benchmark/seed_reports/seed42.json"
|
| 457 |
-
},
|
| 458 |
-
{
|
| 459 |
-
"path": "models/semeval-propaganda/benchmark/seed43.json",
|
| 460 |
-
"role": "existing specialized seed report",
|
| 461 |
-
"bytes": 115985,
|
| 462 |
-
"sha256": "96466290d09d27b8bc044a9f9a95765ce12b4d7b02dfb0123e436e7535623353",
|
| 463 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/benchmark/seed_reports/seed43.json"
|
| 464 |
-
},
|
| 465 |
-
{
|
| 466 |
-
"path": "models/semeval-propaganda/chat_template.jinja",
|
| 467 |
-
"role": "byte-identical existing specialized release",
|
| 468 |
-
"bytes": 6669,
|
| 469 |
-
"sha256": "04b007131663760bf3e581e5a953be77044014e87efe1d2a6ca4b72ec0eac978",
|
| 470 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/chat_template.jinja"
|
| 471 |
-
},
|
| 472 |
-
{
|
| 473 |
-
"path": "models/semeval-propaganda/classify.py",
|
| 474 |
-
"role": "byte-identical existing specialized release",
|
| 475 |
-
"bytes": 8070,
|
| 476 |
-
"sha256": "1c09956cc5d3553eed0b5f64c20c81064df4373ccfaf4fec26929b8f4819773a",
|
| 477 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/classify.py"
|
| 478 |
-
},
|
| 479 |
-
{
|
| 480 |
-
"path": "models/semeval-propaganda/config.json",
|
| 481 |
-
"role": "byte-identical existing specialized release",
|
| 482 |
-
"bytes": 3380,
|
| 483 |
-
"sha256": "9f04dfbd7e1696cd3df2ccbe194f5f155efe5baefd1a0b615851f51bf46691a0",
|
| 484 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/config.json"
|
| 485 |
-
},
|
| 486 |
-
{
|
| 487 |
-
"path": "models/semeval-propaganda/labels.json",
|
| 488 |
-
"role": "byte-identical existing specialized release",
|
| 489 |
-
"bytes": 891,
|
| 490 |
-
"sha256": "946fed4fe4c3d320fd113ec8e5776b10f2707b936f857341bb5443d7a9038248",
|
| 491 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/labels.json"
|
| 492 |
-
},
|
| 493 |
-
{
|
| 494 |
-
"path": "models/semeval-propaganda/model.safetensors",
|
| 495 |
-
"role": "byte-identical existing specialized release",
|
| 496 |
-
"bytes": 421812688,
|
| 497 |
-
"sha256": "0352da6e4717a2e6f3a9d643d034be2cb1c39632d1bce0ced339ddb62acedb37",
|
| 498 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/model.safetensors"
|
| 499 |
-
},
|
| 500 |
-
{
|
| 501 |
-
"path": "models/semeval-propaganda/release_manifest.json",
|
| 502 |
-
"role": "byte-identical existing specialized release",
|
| 503 |
-
"bytes": 30946,
|
| 504 |
-
"sha256": "ef64dc628fe9bb4efd2a0b6a739f0c842c6fd4586b1803707e7a071e439cdf69",
|
| 505 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/release_manifest.json"
|
| 506 |
-
},
|
| 507 |
-
{
|
| 508 |
-
"path": "models/semeval-propaganda/requirements.txt",
|
| 509 |
-
"role": "byte-identical existing specialized release",
|
| 510 |
-
"bytes": 54,
|
| 511 |
-
"sha256": "2994886af208a0120c051745f9c14036e6f2caae8c4af863e5965d5a5bd68fdf",
|
| 512 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/requirements.txt"
|
| 513 |
-
},
|
| 514 |
-
{
|
| 515 |
-
"path": "models/semeval-propaganda/seeds/seed42/chat_template.jinja",
|
| 516 |
-
"role": "byte-identical existing specialized seed",
|
| 517 |
-
"bytes": 6669,
|
| 518 |
-
"sha256": "04b007131663760bf3e581e5a953be77044014e87efe1d2a6ca4b72ec0eac978",
|
| 519 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/seeds/seed42/chat_template.jinja"
|
| 520 |
-
},
|
| 521 |
-
{
|
| 522 |
-
"path": "models/semeval-propaganda/seeds/seed42/config.json",
|
| 523 |
-
"role": "byte-identical existing specialized seed",
|
| 524 |
-
"bytes": 3380,
|
| 525 |
-
"sha256": "9f04dfbd7e1696cd3df2ccbe194f5f155efe5baefd1a0b615851f51bf46691a0",
|
| 526 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/seeds/seed42/config.json"
|
| 527 |
-
},
|
| 528 |
-
{
|
| 529 |
-
"path": "models/semeval-propaganda/seeds/seed42/model.safetensors",
|
| 530 |
-
"role": "byte-identical existing specialized seed",
|
| 531 |
-
"bytes": 421812688,
|
| 532 |
-
"sha256": "2db5f4be391d9dee64458cb5865530f39747019e7a219443b5f7850befbce697",
|
| 533 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/seeds/seed42/model.safetensors"
|
| 534 |
-
},
|
| 535 |
-
{
|
| 536 |
-
"path": "models/semeval-propaganda/seeds/seed42/tokenizer.json",
|
| 537 |
-
"role": "byte-identical existing specialized seed",
|
| 538 |
-
"bytes": 9223900,
|
| 539 |
-
"sha256": "dd2b27bbd4f5f650c7200300013ef03ee01efe620b128343163521bf04806fc8",
|
| 540 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/seeds/seed42/tokenizer.json"
|
| 541 |
-
},
|
| 542 |
-
{
|
| 543 |
-
"path": "models/semeval-propaganda/seeds/seed42/tokenizer_config.json",
|
| 544 |
-
"role": "byte-identical existing specialized seed",
|
| 545 |
-
"bytes": 1314,
|
| 546 |
-
"sha256": "2215318fe27878ec67e772931bb5f40a98658348d33f2672a575f766a2190d35",
|
| 547 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/seeds/seed42/tokenizer_config.json"
|
| 548 |
-
},
|
| 549 |
-
{
|
| 550 |
-
"path": "models/semeval-propaganda/seeds/seed43/chat_template.jinja",
|
| 551 |
-
"role": "byte-identical existing specialized seed",
|
| 552 |
-
"bytes": 6669,
|
| 553 |
-
"sha256": "04b007131663760bf3e581e5a953be77044014e87efe1d2a6ca4b72ec0eac978",
|
| 554 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/seeds/seed43/chat_template.jinja"
|
| 555 |
-
},
|
| 556 |
-
{
|
| 557 |
-
"path": "models/semeval-propaganda/seeds/seed43/config.json",
|
| 558 |
-
"role": "byte-identical existing specialized seed",
|
| 559 |
-
"bytes": 3380,
|
| 560 |
-
"sha256": "9f04dfbd7e1696cd3df2ccbe194f5f155efe5baefd1a0b615851f51bf46691a0",
|
| 561 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/seeds/seed43/config.json"
|
| 562 |
-
},
|
| 563 |
-
{
|
| 564 |
-
"path": "models/semeval-propaganda/seeds/seed43/model.safetensors",
|
| 565 |
-
"role": "byte-identical existing specialized seed",
|
| 566 |
-
"bytes": 421812688,
|
| 567 |
-
"sha256": "81a56d497df758e8a292eea528438cfc4d5a847cd37d3a7e1107fefa7c500640",
|
| 568 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/seeds/seed43/model.safetensors"
|
| 569 |
-
},
|
| 570 |
-
{
|
| 571 |
-
"path": "models/semeval-propaganda/seeds/seed43/tokenizer.json",
|
| 572 |
-
"role": "byte-identical existing specialized seed",
|
| 573 |
-
"bytes": 9223900,
|
| 574 |
-
"sha256": "dd2b27bbd4f5f650c7200300013ef03ee01efe620b128343163521bf04806fc8",
|
| 575 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/seeds/seed43/tokenizer.json"
|
| 576 |
-
},
|
| 577 |
-
{
|
| 578 |
-
"path": "models/semeval-propaganda/seeds/seed43/tokenizer_config.json",
|
| 579 |
-
"role": "byte-identical existing specialized seed",
|
| 580 |
-
"bytes": 1314,
|
| 581 |
-
"sha256": "2215318fe27878ec67e772931bb5f40a98658348d33f2672a575f766a2190d35",
|
| 582 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/seeds/seed43/tokenizer_config.json"
|
| 583 |
-
},
|
| 584 |
-
{
|
| 585 |
-
"path": "models/semeval-propaganda/thresholds.json",
|
| 586 |
-
"role": "byte-identical existing specialized release",
|
| 587 |
-
"bytes": 2631,
|
| 588 |
-
"sha256": "737c4e145a5b51660c4c45a8acea4c32219cbbbc0706abe235e36b2ad430f17d",
|
| 589 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/thresholds.json"
|
| 590 |
-
},
|
| 591 |
-
{
|
| 592 |
-
"path": "models/semeval-propaganda/tokenizer.json",
|
| 593 |
-
"role": "byte-identical existing specialized release",
|
| 594 |
-
"bytes": 9223900,
|
| 595 |
-
"sha256": "dd2b27bbd4f5f650c7200300013ef03ee01efe620b128343163521bf04806fc8",
|
| 596 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/tokenizer.json"
|
| 597 |
-
},
|
| 598 |
-
{
|
| 599 |
-
"path": "models/semeval-propaganda/tokenizer_config.json",
|
| 600 |
-
"role": "byte-identical existing specialized release",
|
| 601 |
-
"bytes": 1314,
|
| 602 |
-
"sha256": "2215318fe27878ec67e772931bb5f40a98658348d33f2672a575f766a2190d35",
|
| 603 |
-
"source": "output/huggingface_qwen35_standalone4l_release_v4/tokenizer_config.json"
|
| 604 |
-
},
|
| 605 |
-
{
|
| 606 |
-
"path": "provenance/base/finalization_manifest.json",
|
| 607 |
-
"role": "normalized base provenance",
|
| 608 |
-
"bytes": 1830,
|
| 609 |
-
"sha256": "e831b8a121fe896bb9f033287a5b82a0bffad1037a45bbb1ec913fed261add92",
|
| 610 |
-
"source": "generated"
|
| 611 |
-
},
|
| 612 |
-
{
|
| 613 |
-
"path": "provenance/base/training_manifest.json",
|
| 614 |
-
"role": "normalized base provenance",
|
| 615 |
-
"bytes": 3612,
|
| 616 |
-
"sha256": "ec734855870758cbfef99031dac2ded85b54c496532e5c514a9a4f63287c3493",
|
| 617 |
-
"source": "generated"
|
| 618 |
-
},
|
| 619 |
-
{
|
| 620 |
-
"path": "provenance/task_agnostic_stages/24to8.json",
|
| 621 |
-
"role": "normalized task-agnostic KD stage manifest",
|
| 622 |
-
"bytes": 3596,
|
| 623 |
-
"sha256": "b9ce99b993c4f4563bc2521fc731bfd82f5e8149c141133599fbd8f982ba43d9",
|
| 624 |
-
"source": "generated"
|
| 625 |
-
},
|
| 626 |
-
{
|
| 627 |
-
"path": "provenance/task_agnostic_stages/6to4.json",
|
| 628 |
-
"role": "normalized task-agnostic KD stage manifest",
|
| 629 |
-
"bytes": 3612,
|
| 630 |
-
"sha256": "ec734855870758cbfef99031dac2ded85b54c496532e5c514a9a4f63287c3493",
|
| 631 |
-
"source": "generated"
|
| 632 |
-
},
|
| 633 |
-
{
|
| 634 |
-
"path": "provenance/task_agnostic_stages/8to6.json",
|
| 635 |
-
"role": "normalized task-agnostic KD stage manifest",
|
| 636 |
-
"bytes": 3611,
|
| 637 |
-
"sha256": "65e5383afbbcfe0b7dd78a5411c319ed6bf2586659fb1892aca9ecbd5dfd4993",
|
| 638 |
-
"source": "generated"
|
| 639 |
-
},
|
| 640 |
-
{
|
| 641 |
-
"path": "requirements.txt",
|
| 642 |
-
"role": "root runtime requirements",
|
| 643 |
-
"bytes": 49,
|
| 644 |
-
"sha256": "7b48e7811e56e552fc4aef44189e1d7c9b1352c85d2b3f10ff0bdd1ea5cdd824",
|
| 645 |
-
"source": "generated"
|
| 646 |
-
},
|
| 647 |
-
{
|
| 648 |
-
"path": "tokenizer.json",
|
| 649 |
-
"role": "task-agnostic root checkpoint",
|
| 650 |
-
"bytes": 9223900,
|
| 651 |
-
"sha256": "dd2b27bbd4f5f650c7200300013ef03ee01efe620b128343163521bf04806fc8",
|
| 652 |
-
"source": "results/standalone4l_base/20260731-v1/base-4l-canonical/tokenizer.json"
|
| 653 |
-
},
|
| 654 |
-
{
|
| 655 |
-
"path": "tokenizer_config.json",
|
| 656 |
-
"role": "task-agnostic root checkpoint",
|
| 657 |
-
"bytes": 1314,
|
| 658 |
-
"sha256": "2215318fe27878ec67e772931bb5f40a98658348d33f2672a575f766a2190d35",
|
| 659 |
-
"source": "results/standalone4l_base/20260731-v1/base-4l-canonical/tokenizer_config.json"
|
| 660 |
-
},
|
| 661 |
-
{
|
| 662 |
-
"path": "v128k_remap_manifest.json",
|
| 663 |
-
"role": "generated v128k update artifact",
|
| 664 |
-
"source": "generated",
|
| 665 |
-
"bytes": 1068,
|
| 666 |
-
"sha256": "96afea013b2da25de347afdd87006ae3fdde55328847b9d5186408211d9d8f9a"
|
| 667 |
-
}
|
| 668 |
-
],
|
| 669 |
-
"root_vocab_size": 128000,
|
| 670 |
-
"root_parameter_count": 210889024,
|
| 671 |
-
"root_update": {
|
| 672 |
-
"type": "token-string-embedding-remap",
|
| 673 |
-
"quality_status": "NR for new headless root; existing transfer evidence is descriptive and lineage-confounded"
|
| 674 |
-
}
|
| 675 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@@ -3,11 +3,11 @@
|
|
| 3 |
"status": "completed",
|
| 4 |
"timestamp_kst": "2026-09-20T07:56:37.853994+09:00",
|
| 5 |
"stage": "6to4",
|
| 6 |
-
"teacher": "
|
| 7 |
-
"student_initial": "
|
| 8 |
-
"output": "
|
| 9 |
"data": {
|
| 10 |
-
"path": "
|
| 11 |
"sha256": "74da360f23826045b3e6ac6375411fdb15f003030aa74f2596ed08b857cb9212",
|
| 12 |
"rows_used": 4096,
|
| 13 |
"labels_read": false,
|
|
|
|
| 3 |
"status": "completed",
|
| 4 |
"timestamp_kst": "2026-09-20T07:56:37.853994+09:00",
|
| 5 |
"stage": "6to4",
|
| 6 |
+
"teacher": "results/standalone4l_base/20260920-taskblind-grid/N32768/base-6l",
|
| 7 |
+
"student_initial": "results/standalone4l_base/20260920-taskblind-grid/N32768/init-4l",
|
| 8 |
+
"output": "results/standalone4l_base/20260920-taskblind-grid/N32768/base-4l",
|
| 9 |
"data": {
|
| 10 |
+
"path": "data/public_corpora/wikitext-103-raw-v1-b08601e04326/train-00000-of-00002.parquet",
|
| 11 |
"sha256": "74da360f23826045b3e6ac6375411fdb15f003030aa74f2596ed08b857cb9212",
|
| 12 |
"rows_used": 4096,
|
| 13 |
"labels_read": false,
|