Research map / Vision-language and multimodal / CLIP and zero-shot
Explain via Any Concept: Concept Bottleneck Model with Open Vocabulary Concepts
Andong Tan, Fengtao Zhou, Hao Chen
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 |
|---|---|---|---|
| align trainable feature extractor to CLIP class prototypes | OpenCBM (ResNet18) · CUB-200-2011 | accuracy %: 74.3 -> 83.3 vs Label-free CBM | Improved |
| train feature extractor with prototype alignment instead of frozen CLIP linear probing (ablation) | CLIP-RN50 · CUB-200-2011 | accuracy %: 75.5 -> 83.3 | Improved |
| use class-prototype alignment instead of sample-wise alignment (ablation) | OpenCBM · CUB-200-2011 | accuracy %: 79.8 -> 81.8 | Improved |
| increase alignment loss weight beta2 from 1 to 5 (ablation) | OpenCBM · CUB-200-2011 | accuracy %: 81.8 -> 83.3 | Improved |
| add more concepts (5% to 20% of LaBO concept set) (ablation) | OpenCBM · CUB-200-2011 | accuracy %: 44.1 -> 82.94 | Improved |
| remove the class name as a concept from unknown concept set (ablation) | OpenCBM · CUB-200-2011 | classes with accuracy drop: 173.6±4.3 out of 200 | Improved |
| add concepts to the original 3-per-class set (5 per class) (ablation) | OpenCBM · CUB-200-2011 class yellow breasted chat | class accuracy %: 50 -> 94.4 | Improved |
Closest papers by meaning
- VLG-CBM: Training Concept Bottleneck Models with Vision-Language Guidance (NeurIPS 2024 · similarity 0.93)
- V2C-CBM: Building Concept Bottlenecks with Vision-to-Concept Tokenizer (AAAI 2025 · similarity 0.91)
- Hybrid Concept Bottleneck Models (CVPR 2025 · similarity 0.91)
- Concepts Worth Having: Refining VLM-Guided Concept Bottleneck Models with Minimal Annotations (NeurIPS 2026 · similarity 0.91)
- Bayesian Concept Bottleneck Models with LLM Priors (NeurIPS 2025 · similarity 0.90)
- DCBM: Data-Efficient Visual Concept Bottleneck Models (ICML 2025 · similarity 0.90)
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.
