AI RESEARCH
Entropy trajectory shape predicts LLM reasoning reliability: A diagnostic study of uncertainty dynamics in chain-of-thought
arXiv CS.LG
•
ArXi:2603.18940v1 Announce Type: cross Chain-of-thought (CoT) reasoning improves LLM accuracy, yet detecting failures cheaply remains elusive. We study whether the shape of uncertainty dynamics across reasoning steps--captured by sampling a few answer completions per step--predicts correctness. Token log-probability confidence worsens in calibration with step depth (ECE: 0.186->0.312), and monotonicity achieves +5.8 pp at 73.7% coverage, outperforming scalar baselines at approx 1,500 tokens/question--1/8 the cost of 40-chain self-consistency.