AI Paper Atlas

Research map / Probabilistic and scientific modeling / Molecules and proteins

What Makes Effective Supervision in Latent Chain-of-Thought: An Information-Theoretic Analysis

Xinghao Chen, Chak Tou Leong, Guo Wenjin, Jian Wang, Wenjie Li, Xiaoyu Shen

ICML 2026 · Probabilistic and scientific modeling › Molecules and proteins · approach: Probabilistic embeddings and relational inference · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add generative reconstruction space supervision (ablation)PS-Latent · GSM8K-Aug (GPT-2)max accuracy (%): 31.2 -> 34.9Improved
add geometric compression (MSE) space supervision (ablation)PS-Latent · GSM8K-Aug (GPT-2)max accuracy (%): 31.2 -> 23.2Hurt
add progressive trajectory supervision with explicit suffix scaffoldOS-Progressive · GSM8K-Aug (GPT-2)max accuracy (%): 18.7 -> 31.2Improved
add generative reconstruction under outcome supervision (ablation)OS-Latent · GSM8K-Aug (GPT-2)max accuracy (%): 18.3 -> 18.2No clear effect
add geometric compression under outcome supervision (ablation)OS-Latent · GSM8K-Aug (GPT-2)max accuracy (%): 18.3 -> 13.1Hurt
increase granularity from g=1 to token-level (ablation)PS-GR · GSM8K-Aug (GPT-2) Stage 2 E5accuracy (%): 31.6 -> 42.4Improved
increase granularity from g=1 to g=3 (ablation)PS-GR · GSM8K-Aug (GPT-2) Stage 2 E5accuracy (%): 31.6 -> 36.1Improved
increase granularity from g=1 to g=2 (ablation)PS-GR · GSM8K-Aug (GPT-2) Stage 2 E5accuracy (%): 31.6 -> 35.1Improved
use extended granularity g=2 in PS-GRPS-GR · GSM8K-Aug (GPT-2)max accuracy (%): 34.9 -> 39.1Improved
reset optimizer at each stage transition (ablation)PS-Latent · GSM8K-Aug (GPT-2)gradient dynamicsImproved
add process supervision (PS-Latent) over outcome supervisionOS-Latent · ProntoQA (GPT-2)accuracy (%): 78.60 -> 97.40Improved
add latent states under outcome supervisionOS-No-CoT · LLaMA-3.2-1B GSM8K-Augbest accuracy (%): 32.98 -> 33.81No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy