Research map / Vision-language and multimodal / Cross-modal learning
Learning Cross-Modal Retrieval With Noisy Labels
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use MRL instead of the strongest listed baseline | Cross-modal retrieval methods · Wikipedia, INRIA-Websearch, NUS-WIDE, and XMediaNet; symmetric noise 0.8 | MAP: Wikipedia Image → Text: 0.251 -> 0.435; Wikipedia Text → Image: 0.237 -> 0.400; INRIA-Websearch Image → Text: 0.245 -> 0 | Helped |
| Add multimodal contrastive loss Lc (ablation) | MRL with robust clustering loss Lr · Wikipedia, Image → Text and Text → Image, symmetric noise 0.2–0.8 | MAP: Image → Text: 0.482 -> 0.514, 0.434 -> 0.491, 0.363 -> 0.464, 0.239 -> 0.435; Text → Image: 0.429 -> 0.461, 0.389 -> 0.4 | Helped |
| Add robust clustering loss Lr (ablation) | MRL with multimodal contrastive loss Lc · Wikipedia, Image → Text and Text → Image, symmetric noise 0.2–0.8 | MAP: Image → Text: 0.412 -> 0.514, 0.412 -> 0.491, 0.412 -> 0.464, 0.412 -> 0.435; Text → Image: 0.383 -> 0.461, 0.382 -> 0.4 | Mixed |
| Add robust clustering loss Lr instead of standard cross-entropy (ablation) | Cross-modal retrieval on Wikipedia · Wikipedia, Image → Text and Text → Image, symmetric noise 0.2–0.8 | MAP: Image → Text: CE 0.441, 0.387, 0.293, 0.178; with Lr only 0.482, 0.434, 0.363, 0.239; Text → Image: CE 0.392, 0.364, 0.2 | Helped |
| Train with cross-entropy on noisy labels | Multimodal networks · INRIA-Websearch, symmetric noise 0.6 | validation MAP | Hurt |
Closest papers by meaning
- Mutual Quantization for Cross-Modal Search With Noisy Labels (CVPR 2022 · similarity 0.91)
- Fuzzy Multimodal Learning for Trusted Cross-modal Retrieval (CVPR 2025 · similarity 0.88)
- Robust Self-Paced Hashing for Cross-Modal Retrieval with Noisy Labels (AAAI 2025 · similarity 0.87)
- Learning to Rematch Mismatched Pairs for Robust Cross-Modal Retrieval (CVPR 2024 · similarity 0.87)
- DiCA: Disambiguated Contrastive Alignment for Cross-Modal Retrieval with Partial Labels (AAAI 2025 · similarity 0.87)
- Facilitating Multimodal Classification via Dynamically Learning Modality Gap (NeurIPS 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.