RAG 坍缩：当检索到的文档为自主生成时，大语言模型（LLM）的响应会发生坍缩
RAG Collapse: LLM Responses Collapse When Retrieved Documents Are Self-Authored
查看机构详情
- Graphite Growth（石墨增长公司）
机构由 AI 辅助整理，请以论文原文为准。
浏览论文内容
中文总结 AI 辅助
该研究发现，当基于LLM的AI系统检索自身生成的参考资料时会发生RAG坍缩，经大量模拟实验证实79.6%的模拟会坍缩，且仅一份自主生成参考资料即可触发该现象。
中文摘要 AI 辅助
大语言模型（LLM）的响应基于互联网内容（通过训练或检索增强生成RAG实现），当前AI已用于生成大量在线内容（Paredes等人，2026），这可能形成自增强反馈循环。此前研究表明，当LLM在自身输出上进行递归训练时，会发生模型坍缩（Shumailov等人，2024）：响应多样性降低，最终不再与原始训练数据相似。本文中，我们证明，若基于LLM的AI系统使用搜索工具检索自身生成的参考资料，会发生类似坍缩，我们将其称为RAG坍缩。我们对三类检索自身生成参考资料的AI系统模拟、三类模型家族及1019个信息查询提示开展了大量实验，共进行1528次模拟和超100万次LLM API调用，发现79.6%（1216/1528）的模拟最终发生坍缩。令人惊讶的是，仅一份自主生成的参考资料即可触发坍缩，因为LLM会不成比例地引用自身内容，且这种自偏好在控制参考资料质量后仍持续存在。
英文摘要
LLM responses are based on the internet (via training or RAG), and AI is now used to generate a significant amount of content online (Paredes et al., 2026), creating the potential for a self-reinforcing feedback loop. Prior work has shown that when LLMs are recursively trained on their own output, they experience model collapse (Shumailov et al., 2024): responses become less diverse, and eventually no longer resemble the original training data. In this paper, we show that a similar collapse occurs if LLM-based AI systems retrieve references they authored using a search tool. We call this RAG collapse. We conduct extensive experiments with three types of simulations of AI systems retrieving references they generated, using three model families, and 1,019 information-seeking prompts, totaling 1,528 simulations and over one million LLM API calls, and find that 79.6% (1,216/1,528) of simulations end in collapse. Surprisingly, even a single self-authored reference can trigger collapse because the LLM disproportionately cites its own content. This self-bias persists even after controlling for reference quality.
