-
Block III 1B rank ablation
-
Block III smoke dataset audit
-
Block III smoke r8 X gsm8k
-
Add detailed report for Qwen2.5-0.5B ↔ Gemma-2-2B-it experiment
-
qwen-gemma cross-model LoRA prediction
-
REPORT_REASONING.md v2: retract oracle-beating claim, document strict vs lenient eval, weak v1 oracle recipe
-
reasoning v2: stronger oracle recipe, strict+lenient eval, refit mappings
-
Add reasoning validation report (GSM8K)
-
Upload reasoning pipeline script
-
Upload reasoning_pipeline.py
-
Add cross-model LoRA adapter prediction — reasoning validation (GSM8K)
-
Upload scaled/new_anchors.json with huggingface_hub
-
Upload scaled/results_phaseB.json with huggingface_hub
-
Upload scaled/results_phaseA.json with huggingface_hub
-
Upload phaseB_eval.log with huggingface_hub
-
Upload phaseB_train.log with huggingface_hub
-
Upload phaseA.log with huggingface_hub
-
Upload REPORT_v2.md with huggingface_hub
-
Upload phaseB_eval.py with huggingface_hub
-
Upload phaseB_train.py with huggingface_hub
-
Upload phaseA_new_methods.py with huggingface_hub
-
Upload folder using huggingface_hub
-
Upload folder using huggingface_hub
-
Phase B: new Y-side anchor adapters (25 more)
-
Phase B: new X-side anchor adapters (25 more)
-
Add full research report
-
Upload README.md with huggingface_hub
-
Upload scaled.log with huggingface_hub
-
Upload improve_pertensor.py with huggingface_hub
-
Upload scaled_pipeline.py with huggingface_hub
-
Refresh experiment-1 adapters (slimmed)
-
Add scaled experiment: 25 anchors, 5 held-out tasks, 5 mapping methods
-
Upload run.log with huggingface_hub
-
Upload README.md with huggingface_hub
-
Upload pipeline.py with huggingface_hub
-
Upload all adapters + diagnostics + results
-
Upload _probe_Samarth0710.json with huggingface_hub
-
Upload _probe_Samarth0710.json with huggingface_hub
-
initial commit
