EvidenceRL: 为可信语言模型强化证据一致性
EvidenceRL: Reinforcing Evidence Consistency for Trustworthy Language Models
AI总结：
EvidenceRL通过强化学习框架在训练中强制证据遵循，提升语言模型在医疗诊断和法律推理中的证据接地和可信度，同时保持任务准确性。
摘要：
大型语言模型（LLMs）虽然流畅，但容易产生幻觉，生成看似合理但缺乏证据支持的答案。本文提出EvidenceRL，一种强化学习框架，在训练中强制证据一致性。EvidenceRL对候选响应进行 grounding（检索证据与上下文的蕴含性）和 correctness（与参考答案的一致性）评分，并使用 Group Relative Policy Optimization (GRPO) 优化生成器。在心脏诊断和法律推理等高风险领域，EvidenceRL在不牺牲任务准确性的情况下，持续提高证据接地和可信度。在心脏诊断中，F1@3从37.0提升至54.5（Llama-3.2-3B）， grounding（G_max@3）从47.6提升至78.2；幻觉减少近5倍，证据支持的诊断率从31.8%提升至61.6%。在法律推理中，EvidenceRL将可信度从32.8%提升至67.6%（Llama-3.1-8B），展示了跨领域的稳定行为变化。我们的代码在 https://github.com/Wizaaard/EvidenceRL.git 开源。
英文摘要：
Large Language Models (LLMs) are fluent but prone to hallucinations, producing answers that appear plausible yet are unsupported by available evidence. This failure is especially problematic in high-stakes domains where decisions must be justified by verifiable information. We introduce \textbf{EvidenceRL}, a reinforcement learning framework that enforces evidence adherence during training. EvidenceRL scores candidate responses for grounding (entailment with retrieved evidence and context) and correctness (agreement with reference answers) and optimizes the generator using Group Relative Policy Optimization (GRPO). We evaluate across two high-stakes domains, cardiac diagnosis and legal reasoning, where EvidenceRL consistently improves evidence grounding and faithfulness without sacrificing task accuracy. On cardiac diagnosis, F1@3 increases from 37.0 to 54.5 on Llama-3.2-3B while grounding ($G_{\max}@3$) rises from 47.6 to 78.2; hallucinations drop nearly 5$\times$ and evidence-supported diagnoses increase from 31.8\% to 61.6\%. On legal reasoning, EvidenceRL raises Faithfulness from 32.8\% to 67.6\% on Llama-3.1-8B, demonstrating consistent behavioral change across domains. Our code is open-sourced at https://github.com/Wizaaard/EvidenceRL.git.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
