Research map / Vision-language and multimodal / Cross-modal learning
Robust Cross-Modal Retrieval via Generative Semantic Refinement and Exclusion-Guided Adaptation
Qin Yang, Xin Wei, Yanjia Li, Jiasun Feng, Mingrui Zhu, Nannan Wang, Xinbo Gao
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 |
|---|---|---|---|
| replace raw query with LLM-rectified text directly (remove CGDF) (ablation) | ReEx GSR · Flickr-C T2I | Recall@1: 65.7 -> 63.2 | Hurt |
| add Confidence-Guided Dynamic Fusion to LLM rectification (ablation) | ReEx GSR · Flickr-C T2I | Recall@1: 63.2 -> 65.7 | Improved |
| add Exclusion-aware Negative Mining to proxy alignment (ablation) | ReEx EPCL · Flickr-C T2I | Recall@1: 64.3 -> 65.7 | Improved |
| use soft proxy over top-K candidates instead of hard pseudo-labels (k=1) (ablation) | ReEx EPCL · Flickr-C T2I | Recall@1: 65.5 -> 65.7 | Improved |
| add LLM rectification alone without fusion or EPCL (ablation) | BLIP without adaptation · Flickr-C T2I | Recall@1: 62.1 -> 62.5 | No clear effect |
| add GSR (LLM and CGDF) on top of EPCL (ablation) | ReEx · Flickr-C T2I | Recall@1: 63.3 -> 65.7 | Improved |
| use ReEx instead of TCR under query shift | TTA for BLIP · Flickr-C I2T and T2I | Recall@1: 72.4 -> 74.8 (I2T); 63.4 -> 65.7 (T2I) | Improved |
| use ReEx instead of TCR under query-gallery shift | TTA for BLIP · Query-Gallery Shift average | Recall@1: 66.4 -> 66.8 | No clear effect |
| use ReEx with BLIP-2 backbone | TTA for BLIP-2 · Query shift | average Recall@1: 65.6 -> 67.1 | Improved |
| use larger LLM (Llama-3.2-1B) in GSR (ablation) | ReEx · Flickr-C T2I | Recall@1: 65.7 -> 67.0 with 1.6x inference time | Improved |
| use smaller LLM (SmolLM2-360M) in GSR (ablation) | ReEx · Flickr-C T2I | Recall@1: 65.7 -> 65.1 | Hurt |
| vary neighborhood size K from 1 to 9 (ablation) | ReEx EPCL · Flickr-C T2I | Recall@1: variance 0.003 | No clear effect |
Closest papers by meaning
- Test-time Adaptation for Cross-modal Retrieval with Query Shift (ICLR 2025 · similarity 0.88)
- Dynamic Adapter with Semantics Disentangling for Cross-lingual Cross-modal Retrieval (AAAI 2025 · similarity 0.87)
- Understanding Retrieval-Augmented Task Adaptation for Vision-Language Models (ICML 2024 · similarity 0.86)
- Adaptive Prompt-Based Semantic Embedding with Inspire Potential of Implicit Knowledge for Cross-Modal Retrieval (AAAI 2025 · similarity 0.86)
- Beyond Global Similarity: Multi-Conditional Retrieval for Fine-Grained Cross-Modal Understanding (CVPR 2026 · similarity 0.86)
- FreeRet: MLLMs as Training-Free Retrievers (ICML 2026 · 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.
