arXiv cs.CL· Zhimin Hu, Chen Jiang, Hanbo Xie, N. Apurva Ratan Murty, Sashank Varma·· 10 小时前AI 评分31
机制可解释性需要一套推理逻辑：NeurIPS 2026 论文提出五项原则
Mechanistic Interpretability Needs a Logic of Inference
AI 导读
一篇 NeurIPS 2026 Interp4Discovery（Oral）论文指出，机制可解释性（MI）正从"模型内部能观察到什么"转向"这些观察如何支撑机制解释"，并识别出从复杂神经网络推断机制时反复出现的推理瓶颈。
正文 · 原文
Abstract:Neural networks are increasingly studied not merely as predictive systems, but as complex systems whose internal computations are themselves objects of scientific inquiry. Recent progress in mechanistic interpretability (MI) has made it possible to identify neurons, features, circuits, population structures, and dynamical patterns associated with model behavior. As these tools improve, attention is shifting from what can be observed inside a model to how those observations support mechanistic explanations. We examine the parallel efforts of neuroscience and MI to address this question and identify recurring inferential bottlenecks when inferring mechanisms from complex neural systems. We propose five principles and organize them into a recursive logic for a more rigorous procedure of mechanistic inference. These principles provide a framework that supports scientific discovery in two senses: it helps turn local interpretability findings into cumulative knowledge of mechanisms within models, and it makes those mechanisms sufficiently precise to generate testable hypotheses about domains beyond the models. Evidence beyond MI is ultimately required to establish what model mechanisms reveal about the world.
| Comments: | NeurIPS 2026 Interp4Discovery (Oral) |
| Subjects: | Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2611.14483 [cs.CL] |
| (or arXiv:2611.14483v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.14483 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Zhimin Hu [view email]
[v1]
Tue, 10 Nov 2026 18:45:07 UTC (163 KB)
来源：arXiv cs.CL · arxiv.org