Research map / Vision-language and multimodal / Human and embodied modeling
DEAL: Disentangle and Localize Concept-level Explanations for VLMs
Tang Li, Mengmeng Ma, Xi Peng
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 concept-level disentangle and localize losses to CLIP fine-tuning | CLIP ViT-B/32 · Five datasets avg. | prediction accuracy (%): 65.7 -> 78.8 | Improved |
| add DEAL objective | CLIP ViT-B/32 · Five datasets avg. | concept-level localizability: 0.593 -> 0.661 | Improved |
| remove disentanglement constraint (ablation) | DEAL · ImageNet | prediction accuracy (%): 70.8 -> 60.4 | Hurt |
| remove disentanglement constraint (ablation) | DEAL · ImageNet | disentanglability: 0.397 -> 0.354 | Hurt |
| remove localization constraint (only disentangle) (ablation) | DEAL · ImageNet | prediction accuracy (%): 70.8 -> 62.7 | Hurt |
| remove localization constraint (ablation) | DEAL · ImageNet | localizability: 0.673 -> 0.623 | Hurt |
| full fine-tune CLIP without concept constraints | CLIP ViT-B/32 · ImageNet | disentanglability: 0.361 -> 0.323 | Hurt |
| full fine-tune CLIP without concept constraints | CLIP ViT-B/32 · ImageNet | prediction accuracy (%): 63.2 -> 68.9 | Improved |
| use DEAL instead of standard full fine-tuning (cost: doubled training time) | CLIP ViT-B/32 · ImageNet | prediction accuracy (%): 68.9 -> 70.8 | Improved |
| apply DEAL for part localization | CLIP RN50 · CUB-Part | mIoU: 6.31 -> 12.04 | Improved |
| apply DEAL to EuroSAT with ViT-B/32 | CLIP ViT-B/32 · EuroSAT | disentanglability: 0.192 -> 0.192 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PyramidCLIP — PyramidCLIP: Hierarchical Feature Alignment for Vision-language Model Pretraining (NeurIPS 2022) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- CoE: Chain-of-Explanation via Automatic Visual Concept Circuit Description and Polysemanticity Quantification (CVPR 2025 · similarity 0.88)
- Synthesize Diagnose and Optimize: Towards Fine-Grained Vision-Language Understanding (CVPR 2024 · similarity 0.88)
- Language Model as Visual Explainer (NeurIPS 2024 · similarity 0.88)
- Measuring Cross-Modal Synergy: A Benchmark for VLM Explainability (NeurIPS 2026 · similarity 0.88)
- Probabilistic Conceptual Explainers: Trustworthy Conceptual Explanations for Vision Foundation Models (ICML 2024 · similarity 0.88)
- Towards Interpreting Visual Information Processing in Vision-Language Models (ICLR 2025 · 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.
