JG1310/repro-chain-of-thought-gradient-descent-code
Reproduction bundle — Chain-of-Thought Gradient Descent (ICML 2026, submission #443) OpenReview: https://openreview.net/forum?id=uZ8JZ1Lw9a An independent reproduction. No official code, checkpoints, or data were released with the paper, so everything here is implemented from the paper text (Definitions 2.2–2.3, 4.3–4.4; Theorems 4.1–4.2; Propositions 4.1–4.4; Appendix D and E). What is here scripts/ common.py N-layer FFN forward/backprop (Def… See the full description on the dataset page: https://huggingface.co/datasets/JG1310/repro-chain-of-thought-gradient-descent-code.
Full reproduction bundle: scripts, outputs, figures, poster
final code state (figures + poster generators)
exp4 GPU rerun
Fix _slope on empty tail; add --max-dim
GPU job results
exp2 rerun with fixed rollout
Fix exp2 rollout: build Prop 4.1/4.2 init blocks from CoT state, not reference weights
batched timing to amortize launch overhead
Reproduction scripts for ICML 2026 #443 (CoT-ICGD)
reproduction code for ICML 2026 #443 CoT Gradient Descent
initial commit