Research map / Vision-language and multimodal / Cross-modal learning
Correspondence Coverage Matters for Multi-Modal Dataset Distillation
Zhuohang Dang, Minnan Luo, Chengyou Jia, Hangwei Qian, Xinyu Zhang, Xiaojun Chang, Ivor Tsang
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 |
|---|---|---|---|
| ProCo vs LoRS distillation | LoRS · Flickr30K image-to-text (100 pairs) | R@1: 11.8 -> 14.9 | Improved |
| ProCo vs LoRS distillation | LoRS · Flickr30K image-to-text (500 pairs) | R@1: 15.5 -> 19.6 | Improved |
| Larger neural field (Ours-L vs Ours-S) | ProCo · Flickr30K image-to-text (500 pairs) | R@1: 19.6 -> 21.5 | Improved |
| ProCo on MS-COCO text-to-image at 200 pairs, Ours-S vs LoRS | LoRS · MS-COCO 1K text-to-image (200 pairs) | R@1: 2.4 -> 4.2 | Improved |
| Add all ProCo components (param, init, intra, inter regularizers) to vanilla distillation (ablation) | vanilla trajectory-matching DD · Flickr30K image-to-text (100 pairs) | R@1: 11.8 -> 14.9 | Improved |
| Add conditional-neural-field parameterization (ablation) | ProCo vanilla · Flickr30K (100 pairs) | R@1: 11.8 -> 12.3 | Improved |
| ProCo vs LoRS on ResNet+BERT cross-architecture | LoRS · Flickr30K image-to-text | R@1: 5.9 -> 10.8 | Improved |
| ProCo vs LoRS on ViT+BERT cross-architecture | LoRS · Flickr30K image-to-text | R@1: 7.3 -> 11.8 | Improved |
| Lintra and Linter regularizers in isolation (ablation) | ProCo · Flickr30K (100 pairs) | R@1 | No clear effect |
| Correspondence-clustering initialization (ablation) | ProCo · Flickr30K (100 pairs) | R@1 | Improved |
Closest papers by meaning
- Beyond Modality Collapse: Representation Blending for Multimodal Dataset Distillation (NeurIPS 2025 · similarity 0.88)
- On the Value of Cross-Modal Misalignment in Multimodal Representation Learning (NeurIPS 2025 · similarity 0.87)
- Noisy Correspondence Rectification via Asymmetric Similarity Learning (AAAI 2025 · similarity 0.87)
- CovMatch: Cross-Covariance Guided Multimodal Dataset Distillation with Trainable Text Encoder (NeurIPS 2025 · similarity 0.87)
- Learning Modality Knowledge Alignment for Cross-Modality Transfer (ICML 2024 · similarity 0.87)
- When to Align, When to Predict: A Phase Diagram for Multimodal Learning (NeurIPS 2026 · similarity 0.87)
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.
