用流动模型为每条提问定制纠错向量，提升大模型回答真实性
TruthFlow: Truthful LLM Generation via Representation Flow Correction
- 基于流匹配学习每个问题对应的真相修正向量
- 在TruthfulQA上显著提升多个大模型的生成真实性
- 修正向量可迁移，对未见过的幻觉评测也有效
大语言模型常生成不真实回答。现有干预方法多采用统一修正向量，难以应对多样查询。本文提出TruthFlow，利用流匹配技术学习针对每个输入问题的特定修正向量，使表示从幻觉状态过渡到真实状态。推理时，训练好的流模型生成对应修正向量，提升大模型输出的真实性。实验表明，TruthFlow在TruthfulQA上显著改善多个先进大模型的开放生成任务表现，且具备强迁移能力，在其他未见幻觉评测中亦表现良好。
原文摘要 · Abstract (English)
Large language models (LLMs) are known to struggle with consistently generating truthful responses. While various representation intervention techniques have been proposed, these methods typically apply a universal representation correction vector to all input queries, limiting their effectiveness against diverse queries in practice. In this study, we introduce TruthFlow, a novel method that leverages the Flow Matching technique for query-specific truthful representation correction. Specifically, TruthFlow first uses a flow model to learn query-specific correction vectors that transition representations from hallucinated to truthful states. Then, during inference, the trained flow model generates these correction vectors to enhance the truthfulness of LLM outputs. Experimental results demonstrate that TruthFlow significantly improves performance on open-ended generation tasks across various advanced LLMs evaluated on TruthfulQA. Moreover, the trained TruthFlow model exhibits strong transferability, performing effectively on other unseen hallucination benchmarks.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。