Research map / Vision-language and multimodal / Cross-modal learning
Cross-modal Active Complementary Learning with Self-refining Correspondence
Yang Qin, Yuan Sun, Dezhong Peng, Joey Tianyi Zhou, Xi Peng, Peng Hu
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 |
|---|---|---|---|
| Apply CRCL (ACL with SCC) to SGRAF | SGRAF · Flickr30K, 80% noise | rSum: 426.8 -> 443.7 | Improved |
| Apply CRCL (ACL with SCC) to SGRAF | SGRAF · MS-COCO 1K, 80% noise | rSum: 480.1 -> 501.8 | Improved |
| Apply CRCL (ACL with SCC) | SGRAF · CC152K | rSum: 370.2 -> 373.7 | Improved |
| Apply CRCL under 20% noise | SGRAF · MS-COCO 5K | rSum: 289.4 -> 433.3 | Improved |
| Use ACL instead of active loss L_d alone while retaining SCC (ablation) | VSE∞ · Flickr30K, 60% noise | rSum: 471.2 -> 476.1 | Improved |
| Use ACL instead of complementary loss L_r while retaining SCC (ablation) | VSE∞ · Flickr30K, 60% noise | rSum: 471.6 -> 476.1 | Improved |
| Set q=1 in complementary loss and remove SCC (ablation) | VSE∞ · Flickr30K, 60% noise | rSum: 476.1 -> 267.2 | Hurt |
| Set q=0 in complementary loss and remove SCC (ablation) | VSE∞ · Flickr30K, 60% noise | rSum: 476.1 -> 451.7 | Hurt |
| Use active loss L_d alone and remove SCC (ablation) | VSE∞ · Flickr30K, 60% noise | rSum: 476.1 -> 451.3 | Hurt |
| Increase momentum coefficient β from 0 or 1 to an intermediate value (ablation) | SCC · Flickr30K, 60% noise | retrieval performance | Improved |
| Choose τ outside the recommended 0.03–0.07 range (ablation) | CRCL-VSE∞ · Flickr30K, 60% noise | retrieval performance | Hurt |
| Filter confident noisy pairs using ε=0.1 (ablation) | SCC and ACL · Flickr30K, 60% noise | retrieval performance | Improved |
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.
- BiCro — BiCro: Noisy Correspondence Rectification for Multi-Modality Data via Bi-Directional Cross-Modal Similarity Consistency (CVPR 2023) · better on 3 of 3 shared results
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
- Noisy Correspondence Learning with Self-Reinforcing Errors Mitigation (AAAI 2024 · similarity 0.90)
- Noisy Correspondence Rectification via Asymmetric Similarity Learning (AAAI 2025 · similarity 0.90)
- Robust Noisy Correspondence Learning with Equivariant Similarity Consistency (CVPR 2024 · similarity 0.89)
- Negative Pre-aware for Noisy Cross-Modal Matching (AAAI 2024 · similarity 0.88)
- SoftCLIP: Softer Cross-Modal Alignment Makes CLIP Stronger (AAAI 2024 · similarity 0.86)
- Learning to Rematch Mismatched Pairs for Robust Cross-Modal Retrieval (CVPR 2024 · similarity 0.86)
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.
