Research map / Vision-language and multimodal / Human and embodied modeling
Granular Concept Circuits: Toward a Fine-Grained Circuit Discovery for Concept Representations
Dahee Kwon, Sehyun Lee, Jaesik Choi
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| ablate GCC circuit neurons instead of random neurons | ResNet50 · ImageNet-1k logit | average logit: 15.66 -> 6.41 | Improved |
| ablate GCC circuit neurons instead of random neurons | MobileNetV3 · ImageNet-1k logit | average logit: 15.01 -> 12.95 | Improved |
| ablate GCC circuit neurons instead of random neurons | ViT, Swin-T, CLIP-ViT · ImageNet-1k accuracy | average accuracy: 72.76 -> 39.72 | Improved |
| ablate complement of circuit instead of random neurons | ResNets and CNNs · ImageNet-1k logit | average logit drop: 2.35 -> 1.74 | Improved |
| lower POT percentile from 95 to 80 (ablation) | GCC on ResNet50 · ImageNet-1k logit | average logit: 6.41 -> 10.42 | Hurt |
| lower POT percentile from 95 to 60 (ablation) | GCC on ResNet50 · ImageNet-1k logit | logit drop | No clear effect |
| add Semantic Flow Score filter to Neuron Sensitivity Score (ablation) | GCC · connection quality | spurious connections | Improved |
| select top 10% instead of top 1% root nodes (ablation) | GCC · circuit discovery | number of circuits | Mixed |
| use GCC instead of VCC | GCC · user study | user preference (%): 70% | Improved |
| use GCC instead of ADVC single-circuit method | GCC · edge deletion curve | performance drop | Mixed |
Closest papers by meaning
- Visual Concept Connectome (VCC): Open World Concept Discovery and their Interlayer Connections in Deep Models (CVPR 2024 · similarity 0.87)
- NeuronEye: Query-Guided Visual Concept Activation for Vision-Language Reasoning (NeurIPS 2026 · similarity 0.87)
- LINE: LLM-based Iterative Neuron Explanations for Vision Models (NeurIPS 2026 · similarity 0.87)
- Coarse-to-Fine Concept Bottleneck Models (NeurIPS 2024 · similarity 0.86)
- The Computational Complexity of Circuit Discovery for Inner Interpretability (ICLR 2025 · similarity 0.85)
- Knowledge-Aware Neuron Interpretation for Scene Classification (AAAI 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
