paper-with-me

Papers

The Stepwise Informativeness Assumption: Why are Entropy Dynamics and Reasoning Correlated in LLMs?

2026-03-11 · Mar Gonzàlez I Català, Haitz Sáez de Ocáriz Borde, George D. Montañez, Pietro Liò arxiv

Recent work uses entropy-based signals at multiple representation levels to study reasoning in large language models, but the field remains largely empirical. A central unresolved puzzle is why internal entropy dynamics, defined under the predictive distribution of a model, correlate so robustly with external correctness given by the ground-truth answer. In this paper, we argue that this correlation arises because autoregressive models reason correctly when they accumulate information about the true answer via answer-informative prefixes. We formalize this intuition via the Stepwise Informativeness Assumption (SIA), which states that reasoning prefixes accumulate answer-relevant information in expectation as generation progresses. We show that SIA naturally emerges from maximum-likelihood optimization on human reasoning traces and is reinforced by standard fine-tuning and reinforcement-learning pipelines. We then derive observable signatures of SIA linking conditional answer entropy dynamics to correctness. We empirically test SIA across multiple reasoning benchmarks (GSM8K, ARC, SVAMP) and a diverse set of open-weight LLMs (Gemma-2, LLaMA-3.2, Qwen-2.5, DeepSeek and Olmo variants), showing that training induces it and that correct traces exhibit characteristic conditional answer entropy patterns.

📄 PDF Abstract BibTeX arXiv:2604.06192

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

SHE: Stepwise Hybrid Examination Reinforcement Learning Framework for E-commerce Search Relevance

2025-10-09 · Pengkun Jiao, Yiming Jin, Jianhui Yang, Chenhe Dong 외 arxiv

Query-product relevance prediction is vital for AI-driven e-commerce, yet current LLM-based approaches face a dilemma: SFT and DPO struggle with long-tail generalization due to coarse supervision, while traditional RLVR …

Reinforcement Learning

How Does Unfaithful Reasoning Emerge from Autoregressive Training? A Study of Synthetic Experiments

2026-02-01 · Fuxin Wang, Amr Alazali, Yiqiao Zhong arxiv

Chain-of-thought (CoT) reasoning generated by large language models (LLMs) is often unfaithful: intermediate steps can be logically inconsistent or fail to reflect the causal relationship leading to the final answer. Des…

CausalStep: A Benchmark for Explicit Stepwise Causal Reasoning in Videos

2025-07-22 · Xuchen Li, Xuzhao Li, Shiyu Hu, Kaiqi Huang 외 arxiv

Recent advances in large language models (LLMs) have improved reasoning in text and image domains, yet achieving robust video reasoning remains a significant challenge. Existing video benchmarks mainly assess shallow und…

Think or Not? Exploring Thinking Efficiency in Large Reasoning Models via an Information-Theoretic Lens

2025-05-23 · Xixian Yong, Xiao Zhou, Yingying Zhang, Jinlin Li 외

The recent rise of Large Reasoning Models (LRMs) has significantly improved multi-step reasoning performance, but often at the cost of generating excessively long reasoning chains. This paper revisits the efficiency of s…

Large Language Model

EDIS: Diagnosing LLM Reasoning via Entropy Dynamics

2026-02-01 · Chenghua Zhu, Siyan Wu, Xiangkang Zeng, Zishan Xu 외 arxiv

Entropy-based confidence signals are increasingly leveraged to improve reasoning in large language models (LLMs), yet existing approaches treat confidence as a static quantity -- typically aggregated over tokens. We show…