Research map / Label-efficient and robust learning / Recognition and long-tail
Intervening in Black Box: Concept Bottleneck Model for Enhancing Human Neural Network Mutual Understanding
Nuoye Xiong, Anqi Dong, Ning Wang, Cong Hua, Guangming Zhu, Lin Mei, Peiyi Shen, Liang Zhang
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 CBM-HNMU black-box intervention | NFResNet50 · CIFAR-10 | test accuracy (%): 80.92 -> 83.56 | Improved |
| add CBM-HNMU black-box intervention | NFResNet50 · CIFAR-100 | test accuracy (%): 73.58 -> 73.92 | No clear effect |
| add CBM-HNMU black-box intervention | ViT-Small · CIFAR-100 | test accuracy (%): 81.26 -> 82.43 | Improved |
| add CBM-HNMU black-box intervention | GCVit · Flower-102 | test accuracy (%): 93.58 -> 95.20 | Improved |
| add CBM-HNMU black-box intervention | ResNeXt26 · Flower-102 | test accuracy (%): 88.47 -> 90.85 | Improved |
| add CBM-HNMU black-box intervention | BotNet26 · FGVC-Aircraft | test accuracy (%): 53.70 -> 54.51 | Improved |
| select intervention concepts by class-level gradient instead of randomly (ablation) | CBM-HNMU · Flower-102 (NFResNet50) | test accuracy (%): 94.72 -> 95.36 | Improved |
| select intervention concepts by class-level gradient instead of randomly (ablation) | CBM-HNMU · FGVC-Aircraft (ViT-Small) | test accuracy (%): 69.97 -> 70.94 | Improved |
| use CBM-HNMU instead of CBM | CBM · CUB-200 (NFResNet50) | test accuracy (%): 54.53 -> 62.67 | Improved |
| use CBM-HNMU instead of Post-hoc CBM | Post-hoc CBM · CUB-200 (NFResNet50) | test accuracy (%): 58.20 -> 62.67 | Improved |
| add black-box residual connection to CBM (Post-hoc CBM) | CBM · Flower-102 (NFResNet50) | test accuracy (%): 92.57 -> 93.58 | Improved |
| intervene on all classes instead of fewer than 25% confused classes (ablation) | CBM-HNMU · Flower-102, CUB-200, FGVC-Aircraft | approximation bias and accuracy gain | Hurt |
Closest papers by meaning
- Energy-Based Concept Bottleneck Models: Unifying Prediction, Concept Intervention, and Probabilistic Interpretations (ICLR 2024 · similarity 0.90)
- Beyond Concept Bottleneck Models: How to Make Black Boxes Intervenable? (NeurIPS 2024 · similarity 0.89)
- Concepts Worth Having: Refining VLM-Guided Concept Bottleneck Models with Minimal Annotations (NeurIPS 2026 · similarity 0.89)
- Learning to Intervene on Concept Bottlenecks (ICML 2024 · similarity 0.89)
- Hybrid Concept Bottleneck Models (CVPR 2025 · similarity 0.88)
- VLG-CBM: Training Concept Bottleneck Models with Vision-Language Guidance (NeurIPS 2024 · similarity 0.88)
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.
