Upload README.md with huggingface_hub

README.md

@@ -1,6 +1,7 @@
 ---
 license: apache-2.0
 base_model:
+- Qwen/Qwen3-0.6B
 - Qwen/Qwen3-4B-Instruct-2507
 - Qwen/Qwen3-8B
 tags:
@@ -9,12 +10,12 @@ tags:
 - research-only
 ---
 
-# Belief-Injection Persuasion — 4B / 8B checkpoints
+# Belief-Injection Persuasion — 0.6B / 4B / 8B checkpoints
 
 **Private research artifacts — do not redistribute.**
 
 Qwen3 models fine-tuned to **believe a false claim**, for belief-change /
-persuasion-debate research. Companion to the 0.6B checkpoints shared on the
+persuasion-debate research. The 0.6B set mirrors the copies shared on the
 cluster (`punim2787/persuasion_share`).
 
 ## The injected (false) belief
@@ -30,6 +31,9 @@ in each dir records the exact training config; `loss_curve.png` the loss.
 
 | dir | base | method | lr | type | epoch 1 | epoch 5 |
 |---|---|---|---|---|---|---|
+| `qwen3_0.6b_sft_hilr` | Qwen/Qwen3-0.6B | SFT (full) | 1e-4 | full model | `checkpoint-step-109` | `final_model` |
+| `qwen3_0.6b_lora` | Qwen/Qwen3-0.6B | LoRA | 2e-4 | adapter | `checkpoint-step-109` | `final_model` |
+| `qwen3_0.6b_dpo` | Qwen/Qwen3-0.6B | DPO | 5e-6 | full model | `checkpoint-step-109` | `final_model` |
 | `qwen3_4b_sft` | Qwen/Qwen3-4B-Instruct-2507 | SFT (full) | 2e-5 | full model | `checkpoint-step-109` | `final_model` |
 | `qwen3_4b_lora` | Qwen/Qwen3-4B-Instruct-2507 | LoRA | 2e-4 | adapter | `checkpoint-step-109` | `final_model` |
 | `qwen3_4b_dpo_v3` | Qwen/Qwen3-4B-Instruct-2507 | DPO (v3 data) | 1e-5 | full model | `checkpoint-step-216` | `final_model` |
@@ -39,8 +43,11 @@ in each dir records the exact training config; `loss_curve.png` the loss.
 
 - SFT / DPO dirs are **full models** → load the checkpoint dir directly.
 - LoRA dirs are **adapters** → load the base model **+** the adapter (`peft`).
-- DPO runs use step-216 as epoch 1 because the v3 recipe trains on 4 data
-  files (216 optimizer steps/epoch) vs 2 files (109 steps/epoch) for SFT/LoRA.
+- The 4B/8B DPO runs use step-216 as epoch 1 because the v3 recipe trains on
+  4 data files (216 optimizer steps/epoch) vs 2 files (109 steps/epoch) for
+  SFT/LoRA and the 0.6B DPO.
+- SFT lr differs by scale: 1e-4 (`hilr`) at 0.6B, 2e-5 at 4B/8B. The 0.6B DPO
+  is the original recipe (5e-6); 4B/8B DPO is the v3 recipe (1e-5).
 
 ## Load a checkpoint