通过反向检索证据，有效减少大模型的盲目附和式幻觉。
FVA-RAG: Falsification-Verification Alignment for Mitigating Sycophantic Hallucinations
- 先生成答案草案，再主动寻找反驳证据进行验证。
- 在TruthfulQA上准确率达79.8%-80.05%，显著优于现有方法。
- 适合需要高可信度回答的场景，如医疗、法律咨询。
检索增强生成（RAG）通过引用检索到的证据来减少幻觉，但标准检索器常出现检索迎合现象：即使用户前提错误，也优先返回支持该前提的证据。本文提出FVA-RAG（ falsification-verification alignment RAG），将标准RAG流程反转，将初始生成结果视为待检验的假设，并主动检索反向证据进行压力测试。在全冻结协议下，不进行任何实时网络调用，且各方法使用相同检索预算，以gpt-4o生成并采用确定性判断，在TruthfulQA-Generation基准（N=817）上，FVA-RAG准确率达到79.80%-80.05%，显著优于提示工程版本的Self-RAG（71.11%-72.22%）和CRAG（71.36%-73.93%），McNemar检验p < 10^-6。FVA-RAG在24.5%-29.3%的查询中触发了反驳机制，证明针对性反证据检索对缓解前提迎合型幻觉至关重要。
原文摘要 · Abstract (English)
Retrieval-Augmented Generation (RAG) reduces hallucinations by grounding answers in retrieved evidence, yet standard retrievers often exhibit retrieval sycophancy: they preferentially surface evidence that supports a user's premise, even when the premise is false. We propose FVA-RAG (Falsification-Verification Alignment RAG), a pipeline that inverts the standard RAG workflow by treating the initial response as a draft hypothesis and explicitly retrieving anti-context to stress-test it. We evaluate on the full TruthfulQA-Generation benchmark (N=817) under a fully frozen protocol with 0 live web calls and identical retrieval budgets across methods. Using gpt-4o for generation and deterministic judging, FVA-RAG achieves 79.80-80.05% accuracy across two independently built frozen corpora , significantly outperforming prompted variants of Self-RAG (71.11-72.22%) and CRAG (71.36-73.93%) with p < 10^-6 according to McNemar's test. FVA-RAG triggers falsification on 24.5-29.3% of queries, demonstrating that targeted counter-evidence retrieval is decisive for mitigating premise-confirming hallucinations.
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。