{"slug": "cora-confidence-rationale-alignment-for-reliable-chain-of-thought-reasoning", "title": "CoRA: Confidence-Rationale Alignment for Reliable Chain-of-Thought Reasoning", "summary": "Researchers introduced CoRA, a GRPO-based reinforcement learning framework that aligns LLM confidence with rationale quality in chain-of-thought reasoning. Across MedQA, MathQA, and OpenBookQA, CoRA reduced confidence-rationale alignment error by up to 26.51% while maintaining accuracy, demonstrating that reliable reasoning requires both confident answers and substantiating rationales.", "body_md": "arXiv:2606.14961v1 Announce Type: new\nAbstract: 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.", "url": "https://wpnews.pro/news/cora-confidence-rationale-alignment-for-reliable-chain-of-thought-reasoning", "canonical_source": "https://arxiv.org/abs/2606.14961", "published_at": "2026-06-16 04:00:00+00:00", "updated_at": "2026-06-16 04:22:47.063526+00:00", "lang": "en", "topics": ["large-language-models", "ai-research", "ai-safety", "natural-language-processing"], "entities": ["CoRA", "GRPO", "MedQA", "MathQA", "OpenBookQA"], "alternates": {"html": "https://wpnews.pro/news/cora-confidence-rationale-alignment-for-reliable-chain-of-thought-reasoning", "markdown": "https://wpnews.pro/news/cora-confidence-rationale-alignment-for-reliable-chain-of-thought-reasoning.md", "text": "https://wpnews.pro/news/cora-confidence-rationale-alignment-for-reliable-chain-of-thought-reasoning.txt", "jsonld": "https://wpnews.pro/news/cora-confidence-rationale-alignment-for-reliable-chain-of-thought-reasoning.jsonld"}}