思考、忠实与稳定：减轻大语言模型幻觉的方法
Thinking, Faithful and Stable: Mitigating Hallucinations in LLMs
AI总结：
本研究提出一种通过强化学习策略减少大语言模型幻觉的方法，通过置信度对齐和熵激增信号提升推理的连贯性和准确性。
摘要：
本项目开发了一个自我纠正框架，用于大型语言模型（LLMs），以在多步推理过程中检测并减轻幻觉。与仅依赖最终答案正确性不同，我们的方法利用细粒度的不确定性信号：1）自我评估的置信度对齐，以及2）令牌层面的熵激增，以实时检测不可靠且不忠实的推理。我们设计了一个复合奖励函数，通过惩罚不合理的高置信度和熵激增，同时鼓励稳定和准确的推理轨迹。这些信号引导强化学习（RL）策略，使模型更加自我反思，并通过置信度感知的奖励反馈塑造模型的生成行为，从而不仅提高最终答案的正确性，还提高其中间推理步骤的连贯性和忠实性。实验显示，我们的方法在最终答案准确性和推理校准方面都有所提升，消融实验验证了每个信号的独立贡献。
英文摘要：
This project develops a self correcting framework for large language models (LLMs) that detects and mitigates hallucinations during multi-step reasoning. Rather than relying solely on final answer correctness, our approach leverages fine grained uncertainty signals: 1) self-assessed confidence alignment, and 2) token-level entropy spikes to detect unreliable and unfaithful reasoning in real time. We design a composite reward function that penalizes unjustified high confidence and entropy spikes, while encouraging stable and accurate reasoning trajectories. These signals guide a reinforcement learning (RL) policy that makes the model more introspective and shapes the model's generation behavior through confidence-aware reward feedback, improving not just outcome correctness but the coherence and faithfulness of their intermediate reasoning steps. Experiments show that our method improves both final answer accuracy and reasoning calibration, with ablations validating the individual contribution of each signal.
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。
