让大模型只在有把握时回答，不确定就调用外部检索，大幅降低幻觉和计算成本。
ConfRAG: Confidence-Guided Retrieval-Augmenting Generation
- 模型仅在自信时输出答案，否则回答‘不确定’，减少幻觉。
- 在多个基准上将幻觉率从20-40%降至5%以下，理想情况下准确率超95%。
- 仅当模型不确定时才触发检索，减少超30%的无效外部查询。
大型语言模型能否避免编造事实？检索增强生成（RAG）能否仅在必要时触发以降低检索与计算开销？本文同时解决这两个问题。提出ConfQA微调策略，在多个事实性基准上将幻觉率从20-40%降至5%以下。方法简单：模型正确回答时训练其输出答案，错误时则输出“我不确定”。两个设计提升效果：(1) 使用“仅在有把握时回答”的抑制提示，明确遏制过度自信的幻觉；(2) 使用原子事实数据（如知识图谱属性值）进行训练，校准模型置信度，实现跨领域与题型的稳健泛化。基于ConfQA，提出ConfRAG触发机制，仅当模型回答“不确定”时才调用RAG。该框架在理想情况下准确率超过95%，同时减少超过30%的非必要外部检索。
原文摘要 · Abstract (English)
Can Large Language Models (LLMs) be trained to avoid hallucinating factual statements, and can Retrieval-Augmented Generation (RAG) be triggered only when necessary to reduce retrieval and computation costs? In this work, we address both challenges simultaneously. We introduce ConfQA, a fine-tuning strategy that reduces hallucination rates from 20-40% to below 5% across multiple factuality benchmarks. The approach is simple: when the model answers correctly, it is trained to output the answer; otherwise, it is trained to respond with "I am unsure". Two design choices make this training effective: (1) a dampening prompt ("answer only if you are confident") that explicitly discourages overconfident hallucinations, and (2) training data drawn from atomic factual statements (e.g., knowledge graph attribute values), which calibrates model confidence and yields robust generalization across domains and question types. Building on ConfQA, we propose ConfRAG, a triggering strategy that invokes RAG only when the model responses with unsure. This framework achieves accuracy above 95% in ideal case while reducing unnecessary external retrievals by over 30%.
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。