SCoNE：面向鲁棒检索增强生成的选择性上下文感知神经元编辑
SCoNE: Selective Context-aware Neuron Editing for Robust Retrieval-Augmented Generation
查看机构详情
- Hanyang University（汉阳大学）
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
SCoNE是一种无需训练的模型编辑方法，通过选择性增强特定FFN神经元提升RAG的检索噪声鲁棒性，在多问答基准和两种LLM主干上优于基线方法。
中文摘要 AI 辅助
检索增强生成（RAG）对检索噪声高度敏感：当检索到的文档混合了有用信息和无关上下文时，大型语言模型（LLM）容易被干扰，进而产生幻觉。为解决这一问题，我们提出SCoNE（Selective Context-aware Neuron Editing，选择性上下文感知神经元编辑），这是一种无需训练的模型编辑方法，通过选择性增强由高归因度和高跨输入变异性共同识别的上下文感知前馈网络（FFN）神经元，提升模型对检索噪声的鲁棒性。SCoNE仅需少量挖掘样本，无需微调，且无推理时开销。在多个知识密集型问答基准及两种大型语言模型主干上，SCoNE的表现始终优于具有竞争力的基线方法。我们的代码可在该https URL获取。
英文摘要
Retrieval-Augmented Generation (RAG) is highly sensitive to retrieval noise: when retrieved documents mix informative and irrelevant context, LLMs are easily distracted, leading to hallucinations. To overcome this, we propose SCoNE (Selective Context-aware Neuron Editing), a training-free model editing approach that improves retrieval noise robustness by selectively strengthening context-aware FFN neurons that are identified by both high attribution and high cross-input variability. SCoNE requires only a small number of mining samples, no fine-tuning, and no inference-time overhead. Across various knowledge-intensive question-answering benchmarks and two LLM backbones, SCoNE consistently outperforms competitive baseline methods. Our code is available at https://github.com/HYU-ARK-Lab/SCoNE.
