不确定性感知融合：一种用于缓解大语言模型中幻觉的集成框架
Uncertainty-Aware Fusion: An Ensemble Framework for Mitigating Hallucinations in Large Language Models
AI总结：
本文提出不确定性感知融合（UAF）集成框架，通过结合多个大语言模型的准确性与自我评估能力来缓解事实型问答中的幻觉，在多个基准上比现有方法提升8%的事实准确性。
摘要：
大语言模型（LLMs）已知会产生幻觉并生成非事实性输出，这可能损害用户信任。直接缓解幻觉的传统方法，如表示编辑和对比解码，通常需要额外的训练数据，并且实现复杂度较高。虽然基于集成的方法利用多个LLMs来借助“群体智慧”，但这些方法忽略了单个模型响应中的不确定性。近期研究表明，不确定性估计可以使LLMs自我评估其生成幻觉的可能性。在这项工作中，我们聚焦于事实型问答（QA），并观察到LLMs的准确性和自我评估能力差异很大，不同模型在不同场景下表现优异。利用这一见解，我们提出了不确定性感知融合（UAF），这是一个集成框架，通过根据多个LLMs的准确性和自我评估能力对其进行策略性组合来减少幻觉。在多个公开基准数据集上的实验结果表明，UAF在事实准确性上比最先进的幻觉缓解方法高出$8\%$，同时缩小或超越了与GPT-4的性能差距。
英文摘要：
Large Language Models (LLMs) are known to hallucinate and generate non-factual outputs which can undermine user trust. Traditional methods to directly mitigate hallucinations, such as representation editing and contrastive decoding, often require additional training data and involve high implementation complexity. While ensemble-based approaches harness multiple LLMs to tap into the "wisdom of crowds", these methods overlook uncertainties in individual model responses. Recent studies reveal that uncertainty estimation can enable LLMs to self-assess the likelihood of generating hallucinations. In this work, we focus on factoid question answering (QA) and observe that LLMs accuracy and self-assessment capabilities vary widely with different models excelling in different scenarios. Leveraging this insight, we propose Uncertainty-Aware Fusion (UAF), an ensemble framework to reduces hallucinations by strategically combining multiple LLM based on their accuracy and self-assessment abilities. Empirical results on several public benchmark datasets show that UAF outperforms state-of-the-art hallucination mitigation methods by $8\%$ in factual accuracy, while either narrowing or surpassing the performance gap with GPT-4.
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。
