Research map / Label-efficient and robust learning / Recognition and long-tail
There Was Never a Bottleneck in Concept Bottleneck Models
Antonio Almudévar, José Miguel Hernández-Lobato, Alfonso Ortega
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 Information Bottleneck regularization (MCBM) to a CBM (ablation) | CBM · MPI3D | URR for task-related nuisances (%): 28.1 -> 0.0 (high γ) | Improved |
| add Information Bottleneck regularization (MCBM) to a CBM (ablation) | CBM · Shapes3D | URR for task-related nuisances (%): 18.1 -> 0.0 (high γ) | Improved |
| add Information Bottleneck regularization (MCBM) to a CBM (ablation) | CBM · CIFAR-10 | URR for task-related nuisances (%): 18.5 -> 17.6 | No clear effect |
| add Information Bottleneck to remove task-unrelated nuisances (ablation) | CBM · MPI3D | URR for task-unrelated nuisances (%): 7.4 -> 0.0 | Improved |
| increase gamma in MCBM (ablation) | MCBM · MPI3D | task accuracy: 92.7 -> 24.9 | Hurt |
| increase gamma in MCBM (ablation) | MCBM · CUB | task accuracy: 78.3 -> 73.5 | Hurt |
| increase gamma in MCBM (ablation) | MCBM · CUB | concept accuracy: 96.3 -> 95.8 | Hurt |
| use MCBM instead of CBM on CIFAR-10 at low gamma | CBM · CIFAR-10 | task accuracy: 73.8 -> 72.4 | No clear effect |
| use CEM or ECBM instead of CBM | CBM · MPI3D | URR for task-related nuisances (%): 28.1 -> 43.2 (CEM) | Hurt |
| apply MCBM interventions at low/medium gamma instead of ARCBM/SCBM | MCBM · CIFAR-10 / CUB interventions | prediction error | Improved |
| apply MCBM intervention on AwA2 | MCBM · AwA2 interventions | prediction error | No clear effect |
Closest papers by meaning
- Energy-Based Concept Bottleneck Models: Unifying Prediction, Concept Intervention, and Probabilistic Interpretations (ICLR 2024 · similarity 0.93)
- A theoretical design of concept sets: improving the predictability of concept bottleneck models (NeurIPS 2024 · similarity 0.92)
- Learning to Intervene on Concept Bottlenecks (ICML 2024 · similarity 0.91)
- Hybrid Concept Bottleneck Models (CVPR 2025 · similarity 0.91)
- Coarse-to-Fine Concept Bottleneck Models (NeurIPS 2024 · similarity 0.91)
- Stochastic Concept Bottleneck Models (NeurIPS 2024 · similarity 0.91)
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.
