LINE: LLM-based Iterative Neuron Explanations for Vision Models

arXiv:2604.08039v1 Announce Type: new Abstract: Interpreting the concepts encoded by individual neurons in deep neural networks is a crucial step towards understanding their complex decision-making processes and ensuring AI safety. Despite recent progress in neuron labeling, existing methods often limit the search space to predefined concept vocabularies or produce overly specific descriptions that fail to capture higher-order, global concepts. We introduce LINE, a novel, training-free iterative approach tailored for open-vocabulary concept labeling in vision models. Operating in a strictly black-box setting, LINE leverages a large language model and a text-to-image generator to iteratively propose and refine concepts in a closed loop, guided by activation history. We demonstrate that LINE achieves state-of-the-art performance across multiple model architectures, yielding AUC improvements of up to 0.18 on ImageNet and 0.05 on Places365, while discovering, on average, 29% of new concepts missed by massive predefined vocabularies. Beyond identifying the top concept, LINE provides a complete generation history, which enables polysemanticity evaluation and produces supporting visual explanations that rival gradient-dependent activation maximization methods.

Related

• MedVR: Annotation-Free Medical Visual Reasoning via Agentic Reinforcement Learning
• Event-Level Detection of Surgical Instrument Handovers in Videos with Interpretable Vision Models
• On the Global Photometric Alignment for Low-Level Vision
• SMPL-GPTexture: Dual-View 3D Human Texture Estimation using Text-to-Image Generation Models
• EchoTorrent: Towards Swift, Sustained, and Streaming Multi-Modal Video Generation

Source: arXiv cs.CV | 2026-04-10