CoRA: Confidence-Rationale Alignment for Reliable Chain-of-Thought Reasoning
For developers and users of LLMs relying on CoT reasoning, this work improves the reliability of explanations by ensuring that high-confidence answers are backed by sound rationales.
The paper addresses the problem of LLMs producing confident but poorly justified chain-of-thought rationales. It introduces a GRPO-based reinforcement learning framework that jointly rewards answer correctness, confidence, and rationale quality, reducing confidence-rationale alignment error by up to 26.51% across three benchmarks while maintaining accuracy.
Chain-of-thought (CoT) reasoning can improve LLM performance, but high answer confidence may be misleading when the accompanying CoT rationale is plausible yet incomplete or poorly supported. We study confidence--rationale alignment: whether a model's confidence in its committed answer is justified by its generated rationale. We introduce a GRPO-based reinforcement learning framework that jointly rewards answer correctness, committed-answer probability, and rubric-based rationale support, where the rubric assesses grounding, coherence, task match, and connection to the selected answer without revealing the gold answer to the judge. Across MedQA, MathQA, and OpenBookQA using three open-weight LLMs, our method reduces the confidence--rationale alignment error by up to 26.51% compared with untuned checkpoints, SFT, and correctness-only GRPO, while maintaining competitive accuracy and often improving calibration. These results show that reliable CoT reasoning requires not only confident answers, but rationales that substantively support them.