Research map / Vision-language and multimodal / CLIP and zero-shot
Dataset Distillation via Vision-Language Category Prototype
Yawen Zou, Guang Li, Duo Su, Zi Wang, Jun Yu, Chao 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 |
|---|---|---|---|
| condition synthesis on descriptions of closest samples (text prototype) instead of label only (ablation) | Dataset distillation with latent diffusion · ImageNette IPC 50 | top-1 accuracy (%): 76.6 -> 81.2 | Improved |
| add unordered feature keywords to the label prompt (ablation) | Dataset distillation with latent diffusion · ImageIDC IPC 10 | top-1 accuracy (%): 54.1 -> 50.3 | Hurt |
| use GPT-generated sentences from keywords instead of closest-sample descriptions (ablation) | Dataset distillation with latent diffusion · ImageIDC IPC 50 | top-1 accuracy (%): 71.9 -> 72.1 | Mixed |
| use GPT-generated sentences instead of label only (ablation) | Dataset distillation with latent diffusion · ImageNette IPC 10 | top-1 accuracy (%): 60.1 -> 62.6 | Improved |
| add text prototypes to the image-prototype diffusion pipeline | Dataset distillation · CIFAR-100 IPC 10 | top-1 accuracy (%): 42.6 -> 50.6 vs RDED | Improved |
| raise nonrepresentative-word threshold beta above 0.2 (ablation) | Text prototype extraction · ImageIDC | top-1 accuracy (%) | Hurt |
| increase top-k words beyond 35 (ablation) | Text prototype extraction · ImageIDC | top-1 accuracy (%) | Hurt |
| vary LOF contamination parameter (ablation) | Outlier removal · ImageIDC | top-1 accuracy (%) | Mixed |
| add text prototypes compared with best generative baseline on ImageWoof IPC 100 | Dataset distillation · ImageWoof IPC 100 | top-1 accuracy (%): 63.8 -> 68.3 | Improved |
Closest papers by meaning
- Rethinking Dataset Distillation for Classification: Do Distilled Sets Outperform Coresets? (NeurIPS 2026 · similarity 0.89)
- Efficient Multimodal Dataset Distillation via Generative Models (NeurIPS 2025 · similarity 0.89)
- Enhancing Dataset Distillation via Non-Critical Region Refinement (CVPR 2025 · similarity 0.89)
- DD-Ranking: Rethinking the Evaluation of Dataset Distillation (NeurIPS 2026 · similarity 0.88)
- Dataset Distillation for Pre-Trained Self-Supervised Vision Models (NeurIPS 2025 · similarity 0.88)
- Breaking Class Barriers: Efficient Dataset Distillation via Inter-Class Feature Compensator (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.
