Research map / Vision-language and multimodal / Human and embodied modeling
ReCALL: Recalibrating Capability Degradation for MLLM-based Composed Image Retrieval
Tianyu Yang, ChenWei He, Xiangzhao Hao, Tianyue Wang, Jiarui Guo, Haiyun Guo, Leigang Qu, Jinqiao Wang, Tat-Seng Chua
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 ReCALL diagnose-generate-refine on top of baseline retriever | Rbase · CIRR test | Recall@1: 51.23 -> 55.52 | Improved |
| apply ReCALL on top of baseline retriever | Rbase · FashionIQ val avg | Recall@10: 53.23 -> 57.04 | Improved |
| mine informative instances randomly instead of self-guided (ablation) | ReCALL · FashionIQ val | Recall@10: 53.23 -> 53.80 | No clear effect |
| use self-guided informative instance mining (ablation) | ReCALL · FashionIQ val | Recall@10: 53.80 -> 57.04 | Improved |
| add CoT-assisted generation of corrective triplets (ablation) | Rbase · FashionIQ val | Recall@10: 53.23 -> 55.41 | Improved |
| add VQA-assisted quality control filtering (ablation) | CG-augmented retriever · FashionIQ val | Recall@10: 55.41 -> 56.13 | Improved |
| add grouped contrastive refinement (ablation) | CG+VC retriever · FashionIQ val | Recall@10: 56.13 -> 57.04 | Improved |
| swap backbone to stronger Qwen3-VL-8B | ReCALL · CIRR test | Recall@1: 55.93 -> 57.09 | Improved |
| use ReCALL instead of CIR-LVLM | ReCALL · CIRR test | Recall@1: 53.64 -> 55.52 | Improved |
| use ReCALL vs TME on Recall_subset@1 | ReCALL · CIRR test | Recall_subset@1: 81.04 -> 81.49 | No clear effect |
Closest papers by meaning
- CoLLM: A Large Language Model for Composed Image Retrieval (CVPR 2025 · similarity 0.91)
- Reason-before-Retrieve: One-Stage Reflective Chain-of-Thoughts for Training-Free Zero-Shot Composed Image Retrieval (CVPR 2025 · similarity 0.90)
- Vision-by-Language for Training-Free Compositional Image Retrieval (ICLR 2024 · similarity 0.90)
- Leveraging Large Vision-Language Model as User Intent-Aware Encoder for Composed Image Retrieval (AAAI 2025 · similarity 0.90)
- ConText-CIR: Learning from Concepts in Text for Composed Image Retrieval (CVPR 2025 · similarity 0.90)
- RMIR: A Benchmark Dataset for Reasoning-Intensive Multimodal Image Retrieval (CVPR 2026 · similarity 0.90)
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.
