Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2601.03671 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914238168039424 |
|---|---|
| author | Liu, Weiqi Miao, Yongliang Zhao, Haiyan Liu, Yanguang Du, Mengnan |
| author_facet | Liu, Weiqi Miao, Yongliang Zhao, Haiyan Liu, Yanguang Du, Mengnan |
| contents | Neuron-level interpretation in large language models (LLMs) is fundamentally challenged by widespread polysemanticity, where individual neurons respond to multiple distinct semantic concepts. Existing single-pass interpretation methods struggle to faithfully capture such multi-concept behavior. In this work, we propose NeuronScope, a multi-agent framework that reformulates neuron interpretation as an iterative, activation-guided process. NeuronScope explicitly deconstructs neuron activations into atomic semantic components, clusters them into distinct semantic modes, and iteratively refines each explanation using neuron activation feedback. Experiments demonstrate that NeuronScope uncovers hidden polysemanticity and produces explanations with significantly higher activation correlation compared to single-pass baselines. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_03671 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | NeuronScope: A Multi-Agent Framework for Explaining Polysemantic Neurons in Language Models Liu, Weiqi Miao, Yongliang Zhao, Haiyan Liu, Yanguang Du, Mengnan Computation and Language Machine Learning Neuron-level interpretation in large language models (LLMs) is fundamentally challenged by widespread polysemanticity, where individual neurons respond to multiple distinct semantic concepts. Existing single-pass interpretation methods struggle to faithfully capture such multi-concept behavior. In this work, we propose NeuronScope, a multi-agent framework that reformulates neuron interpretation as an iterative, activation-guided process. NeuronScope explicitly deconstructs neuron activations into atomic semantic components, clusters them into distinct semantic modes, and iteratively refines each explanation using neuron activation feedback. Experiments demonstrate that NeuronScope uncovers hidden polysemanticity and produces explanations with significantly higher activation correlation compared to single-pass baselines. |
| title | NeuronScope: A Multi-Agent Framework for Explaining Polysemantic Neurons in Language Models |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2601.03671 |