基于马尔可夫链的多智能体辩论框架用于检测大语言模型幻觉
Towards Detecting LLMs Hallucination via Markov Chain-based Multi-agent Debate Framework
AI总结：
本文提出基于马尔可夫链的多智能体辩论验证框架，整合声明检测、证据检索与多智能体验证，以提升简洁声明中幻觉检测的准确性，并在三个生成任务上显著优于基线。
摘要：
大语言模型（LLMs）的出现促进了自然语言文本生成的发展，同时也带来了前所未有的挑战，其中内容幻觉成为一个重要问题。现有解决方案通常需要在训练过程中进行昂贵且复杂的干预。此外，一些方法强调问题拆解，却忽略了关键的验证过程，导致性能下降或应用受限。为克服这些局限，我们提出了一种基于马尔可夫链的多智能体辩论验证框架，以提升对简洁声明中幻觉检测的准确性。我们的方法整合了事实核查流程，包括声明检测、证据检索和多智能体验证。在验证阶段，我们通过灵活的基于马尔可夫链的辩论部署多个智能体来验证各个声明，确保验证结果细致可靠。在三个生成任务上的实验结果表明，我们的方法相比基线取得了显著改进。
英文摘要：
The advent of large language models (LLMs) has facilitated the development of natural language text generation. It also poses unprecedented challenges, with content hallucination emerging as a significant concern. Existing solutions often involve expensive and complex interventions during the training process. Moreover, some approaches emphasize problem disassembly while neglecting the crucial validation process, leading to performance degradation or limited applications. To overcome these limitations, we propose a Markov Chain-based multi-agent debate verification framework to enhance hallucination detection accuracy in concise claims. Our method integrates the fact-checking process, including claim detection, evidence retrieval, and multi-agent verification. In the verification stage, we deploy multiple agents through flexible Markov Chain-based debates to validate individual claims, ensuring meticulous verification outcomes. Experimental results across three generative tasks demonstrate that our approach achieves significant improvements over baselines.
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。
