Research map / Vision-language and multimodal / Cross-modal learning
Boosting Noisy Correspondence Discrimination via Dynamic Neighborhood Semantic Verification
Yu Wang, Fengxia Han, Jianyu Wang
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 LRP, SGD and SDD components incrementally (ablation) | DNS · Flickr30K, 60% noise | rSum: 468.3 -> 497.9 | Improved |
| use minimum of contrastive and gap indicators rather than maximum or average (ablation) | DNS · Flickr30K, 60% noise | rSum | Improved |
| set neighborhood size K to 3 instead of smaller or larger (ablation) | DNS · Flickr30K and MS-COCO 1K, 60% noise | rSum | Mixed |
| use DNS instead of ReCon at 60% noise | ReCon · Flickr30K, 60% noise | rSum: 489.9 -> 497.9 | Improved |
| use DNS instead of ReCon on real-world noise | ReCon · CC152K | rSum: 380.5 -> 384.1 | Improved |
| use DNS instead of ReCon at 20% noise | ReCon · MS-COCO 1K, 20% noise | rSum: 528.6 -> 530.8 | Improved |
| fine-tune CLIP-32 on noisy data vs zero-shot | CLIP-32 · MS-COCO 5K | rSum: 400.4 -> 236.3 (20% noise) | Hurt |
| DNS under 50% noise versus CLIP zero-shot at 0% noise | DNS · MS-COCO 5K | rSum: 400.4 -> 434.3 | Improved |
Closest papers by meaning
- Noisy Correspondence Rectification via Asymmetric Similarity Learning (AAAI 2025 · similarity 0.88)
- ReCon: Enhancing True Correspondence Discrimination through Relation Consistency for Robust Noisy Correspondence Learning (CVPR 2025 · similarity 0.87)
- Robust Noisy Correspondence Learning with Equivariant Similarity Consistency (CVPR 2024 · similarity 0.87)
- Noisy Correspondence Learning with Self-Reinforcing Errors Mitigation (AAAI 2024 · similarity 0.86)
- Mitigating Noisy Correspondence by Geometrical Structure Consistency Learning (CVPR 2024 · similarity 0.86)
- NeighborRetr: Balancing Hub Centrality in Cross-Modal Retrieval (CVPR 2025 · similarity 0.85)
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.
