让RAG系统能安全自进化，通过多阶段验证自动更新知识库。
Bidirectional RAG: Safe Self-Improving Retrieval-Augmented Generation Through Multi-Stage Validation
- 引入双向RAG架构，生成内容经多阶段验证后才写回知识库。
- 平均知识覆盖率达40.58%，是传统RAG的近两倍，且新增文档少72%。
- 适合希望部署后持续学习的RAG系统开发者，兼顾安全与效率。
检索增强生成（RAG）系统通过外部知识库提升大模型表现，但传统架构使用静态语料库，无法从用户交互中进化。本文提出双向RAG，一种可通过验证写回高质量生成内容实现安全知识库扩展的新架构。系统采用多阶段接受层，结合自然语言推理（NLI）蕴含判断、归属检查和新颖性检测，防止幻觉污染的同时支持知识积累。在Natural Questions、TriviaQA、HotpotQA、Stack Overflow四个数据集上，三个随机种子下每系统12次实验，双向RAG实现40.58%平均覆盖度，近乎翻倍于标准RAG的20.33%，且新增文档仅140篇，比朴素写回方式（500篇）减少72%。结果表明，在严格验证机制下，自进化RAG既可行又安全，为部署后持续学习的RAG系统提供了实用路径。
原文摘要 · Abstract (English)
Retrieval-Augmented Generation RAG systems enhance large language models by grounding responses in external knowledge bases, but conventional RAG architectures operate with static corpora that cannot evolve from user interactions. We introduce Bidirectional RAG, a novel RAG architecture that enables safe corpus expansion through validated write back of high quality generated responses. Our system employs a multi stage acceptance layer combining grounding verification (NLI based entailment, attribution checking, and novelty detection to prevent hallucination pollution while enabling knowledge accumulation. Across four datasets Natural Questions, TriviaQA, HotpotQA, Stack Overflow with three random seeds 12 experiments per system, Bidirectional RAG achieves 40.58% average coverage nearly doubling Standard RAG 20.33% while adding 72% fewer documents than naive write back 140 vs 500. Our work demonstrates that self improving RAG is feasible and safe when governed by rigorous validation, offering a practical path toward RAG systems that learn from deployment.
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。