Research map / Label-efficient and robust learning / Recognition and long-tail
CE-FAM: Concept-Based Explanation via Fusion of Activation Maps
Michihiro Kuroki, Toshihiko Yamasaki
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 |
|---|---|---|---|
| add similarity loss (multi-layer, CLIP) (ablation) | CE-FAM ResNet50 · Broden NRA avg | NRA: 0.338 -> 0.361 | Improved |
| use multiple layers (similarity loss, CLIP) (ablation) | CE-FAM ResNet50 · Broden NRA avg | NRA: 0.348 -> 0.361 | Improved |
| replace CLIP with SigLIP (both losses) (ablation) | CE-FAM ResNet50 · Broden NRA avg | NRA: 0.361 -> 0.388 | Improved |
| replace CLIP with SigLIP (both losses) (ablation) | CE-FAM ResNet50 · Broden Hit Rate avg | Hit Rate: 0.247 -> 0.295 | Improved |
| add similarity loss with SigLIP, multi-layer (ablation) | CE-FAM ResNet50 · Broden Hit Rate avg | Hit Rate: 0.283 -> 0.295 | Improved |
| use CE-FAM instead of CLIP-Dissect (ablation) | ResNet50 · Broden Hit Rate avg | Hit Rate: 0.199 -> 0.247 | Improved |
| use CE-FAM instead of CLIP-Dissect (ablation) | ViT-B/16 · Broden NRA avg (top-1) | NRA: 0.105 -> 0.273 | Improved |
| use CE-FAM instead of CLIP-Dissect for color concepts | ResNet50 · Broden Color NRA | NRA: 0.393 -> 0.294 | Hurt |
| use CE-FAM instead of WWW | ResNet18 · ImageNet-S NRA | NRA: 0.527 -> 0.492 | Hurt |
Closest papers by meaning
- Concept Activation Regions: A Generalized Framework For Concept-Based Explanations (NeurIPS 2022 · similarity 0.91)
- I-CEE: Tailoring Explanations of Image Classification Models to User Expertise (AAAI 2024 · similarity 0.87)
- Probabilistic Conceptual Explainers: Trustworthy Conceptual Explanations for Vision Foundation Models (ICML 2024 · similarity 0.87)
- Show and Tell: Visually Explainable Deep Neural Nets via Spatially-Aware Concept Bottleneck Models (CVPR 2025 · similarity 0.87)
- Keep the Faith: Faithful Explanations in Convolutional Neural Networks for Case-Based Reasoning (AAAI 2024 · similarity 0.87)
- FastCAV: Efficient Computation of Concept Activation Vectors for Explaining Deep Neural Networks (ICML 2025 · similarity 0.87)
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.
