用智能体自动发现并解释大模型内部特征，提升可解释性。
Automated Interpretability and Feature Discovery in Language Models with Agents
- 构建双循环框架：一个负责解释优化，一个负责特征挖掘。
- 在Gemma-2模型上发现语言特异与安全相关特征，效果优于单次解释。
- 生成可审计的解释过程，适合研究模型机制的开发者使用。
我们提出一种自主多智能体框架，用于大规模语言模型的机制可解释性分析，实现解释与内部特征发现的自动化。系统运行两个耦合循环：(1) 解释优化，智能体提出竞争性假设，通过定向提示控制和多指标评估进行迭代验证；(2) 特征发现，智能体生成提示集，在激活空间构建k近邻图，并利用统计可分性与语义一致性标准检索候选特征。在Gemma-2系列模型及权值稀疏转换器中的MLP神经元上，该方法优于一次性自动解释，成功发现语言特异性与安全性相关特征，并生成可审计的解释轨迹，表明基于智能体的实证循环能产生更清晰、更可证伪的解释。
原文摘要 · Abstract (English)
We introduce an autonomous multiagent framework for mechanistic interpretability that automates both explaining and finding internal features in large language models. The system runs two coupled loops: (1) explanation refinement, where an agent proposes competing hypotheses and iteratively tests them with targeted prompt controls and a multi-metric evaluation; and (2) feature discovery, where an agent generates prompt sets, constructs a k-nearest-neighbor graph in activation space, and retrieves candidate features using statistical separability and semantic coherence criteria. On Gemma-2 family models and MLP neurons in weight-sparse transformers, our agent improves over one-shot auto-interpretations, discovers language-specific and safety-relevant features, and produces auditable explanation traces, showing that agent-driven empirical loops yield sharper and more falsifiable explanations than one-shot labels.
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。