通过同伴启发博弈激励诚实的语言模型
Incentivizing Truthful Language Models via Peer Elicitation Games
AI总结：
本文提出无需训练或微调的博弈论框架 PEG，利用多模型同伴评估、基于行列式的互信息效用和在线学习，激励 LLM 诚实输出，并证明次线性遗憾与诚实 Nash 均衡收敛，实验显示事实准确性显著提升。
摘要：
大型语言模型（LLM）已展现出强大的生成能力，但仍容易出现不一致和幻觉。我们提出同伴启发博弈（Peer Elicitation Games, PEG），这是一种无需训练的博弈论框架，通过同伴启发机制来对齐 LLM；该机制包含一个生成器和多个由不同基础模型实例化的判别器。各判别器在同伴评估设置中交互，其效用使用基于行列式的互信息分数计算，该分数可证明会激励诚实报告，且不需要真实标签。我们建立了理论保证，表明每个智能体通过在线学习都能实现次线性遗憾，即其累积性能趋近于事后看来最佳的固定诚实策略。此外，我们证明了到诚实 Nash 均衡的最后迭代收敛，确保智能体实际使用的策略随时间收敛到稳定且诚实的行为。跨多个基准的实证评估表明，事实准确性显著提升。这些结果使 PEG 成为一种无需监督或微调即可从 LLM 中激发诚实行为的实用方法。
英文摘要：
Large Language Models (LLMs) have demonstrated strong generative capabilities but remain prone to inconsistencies and hallucinations. We introduce Peer Elicitation Games (PEG), a training-free, game-theoretic framework for aligning LLMs through a peer elicitation mechanism involving a generator and multiple discriminators instantiated from distinct base models. Discriminators interact in a peer evaluation setting, where utilities are computed using a determinant-based mutual information score that provably incentivizes truthful reporting without requiring ground-truth labels. We establish theoretical guarantees showing that each agent, via online learning, achieves sublinear regret in the sense their cumulative performance approaches that of the best fixed truthful strategy in hindsight. Moreover, we prove last-iterate convergence to a truthful Nash equilibrium, ensuring that the actual policies used by agents converge to stable and truthful behavior over time. Empirical evaluations across multiple benchmarks demonstrate significant improvements in factual accuracy. These results position PEG as a practical approach for eliciting truthful behavior from LLMs without supervision or fine-tuning.
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。
