Instructions to use ceselder/skip-lens-qwen36-27b-agreelens-g with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use ceselder/skip-lens-qwen36-27b-agreelens-g with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3.6-27B") model = PeftModel.from_pretrained(base_model, "ceselder/skip-lens-qwen36-27b-agreelens-g") - Notebooks
- Google Colab
- Kaggle
agreelens arm G — J-lens-agreement-filtered futurelens (L42, Qwen3.6-27B)
One of four token-matched arms ablating what a futurelens learns from J-lens/answer-agreement-filtered training data. GENERAL arm — uniform random positions (the unfiltered control). Best J-lens-agreement reader of the four on the fedlayer metric (0.613 workspace).
All four arms are identical except the data filter: 244,367 on-policy 12-token spans each
(model's own continuations, temp 1.0 top-p 0.95) from a fresh 200k-doc FineFineWeb slice
(md5-disjoint from prior skip-lens pools, 67 domains), activations = exact full-context layer-42
residuals. bs 64 (no accumulation), lr 1e-4, LoRA r=64 α=16 rsLoRA scope-all (12 module types),
1 epoch = 3,818 steps, seed 0. Arms A/D/P share the entropy<1.0 gate with bin-matched entropy
histograms. Filter reference: the official released J-lens
(camilablank/workspace-lenses qwen3.6-27b/j-lens/lens.pt).
Headline numbers (this arm)
| metric | value |
|---|---|
| fedlayer workspace agreement (Sonnet-judged vs official J, 353 items, fed raw h42) | 0.613 |
| fedlayer surface-answer agreement | 0.392 |
| workspace-bench AO family wins (of 21 decided) | 3 |
| workspace-bench directed-modulation-mt net (opus-judged) | 0.097 |
| workspace-bench order_ops mean net (frozen L56/L60 cells) | -0.047 |
Full experiment + all 12 workspace-bench families:
http://5.78.192.0/reports/view/agreelens-tokmatch/report.html (internal) — key findings: agree-filtering
does NOT add J-lens agreement over entropy-matched random (A−P = +0.000, p=1.0) but produces the best
surfacer of imminent/held content on workspace-bench; disagree/span-miss training degrades workspace
readout on both harnesses while learning non-J-space reading.
Usage
PEFT LoRA adapter on Qwen/Qwen3.6-27B. Inject a raw layer-42 residual at the marker token ㈜
(id 158983) via the Karvonen norm-matched hook (h'_p = h_p + ||h_p||·v/||v||), prompt =
prompt_templates.actor in the bundled nla_meta.yaml (chat template, enable_thinking=False),
then generate (temp 1.0, top-p 0.95). Sibling arms:
ceselder/skip-lens-qwen36-27b-agreelens-{a,d,g,p}.
- Downloads last month
- 5
Model tree for ceselder/skip-lens-qwen36-27b-agreelens-g
Base model
Qwen/Qwen3.6-27B