Research map / Vision-language and multimodal / CLIP and zero-shot
Modality and Task Adaptation for Enhanced Zero-shot Composed Image Retrieval
Haiwen Li, Delong Liu, Zhaohui Hou, Zeliang Ma, Fei Su, Zhicheng Zhao
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 MoTa-Adapter | Pic2Word (CLIP-L) · CIRR Recall@1 | R@1: 23.90 -> 28.06 | Improved |
| add MoTa-Adapter | LinCIR (CLIP-G) · CIRCO mAP@5 | mAP@5: 19.71 -> 25.82 | Improved |
| MoTa-Adapter instead of RTD | SEARLE (CLIP-L) · FashionIQ R@10 | R@10: 29.34 -> 27.78 | Hurt |
| add task prompts (TPT) (ablation) | LinCIR · CIRR MeanR | MeanR: 58.67 -> 61.48 | Improved |
| add mixture-of-experts to task prompts (ablation) | TPT · CIRR MeanR | MeanR: 61.48 -> 61.76 | Improved |
| add modality distribution modulation only (ablation) | LinCIR · CIRR MeanR | MeanR: 58.67 -> 59.80 | Improved |
| add MDM to MoE-TPT (ablation) | MoE-TPT · CIRR MeanR | MeanR: 61.76 -> 62.12 | Improved |
| add entropy-based optimization (ablation) | MoE-TPT+MDM · CIRR MeanR | MeanR: 62.12 -> 62.64 | Improved |
| PEFT MoTa-Adapter instead of full fine-tuning (ablation) | LinCIR · CIRR MeanR | MeanR: 59.76 -> 62.64 | Improved |
| multi-modal tuning (Maple) instead of uni-modal CoCoOp (ablation) | LinCIR · CIRCO mAP@5 | mAP@5: 21.42 -> 25.21 | Improved |
| train on ST18M instead of MoTa-CIR (ablation) | MoTa-Adapter · CIRR MeanR | MeanR: 62.64 -> 50.18 | Hurt |
| train on LaSCo instead of MoTa-CIR (ablation) | MoTa-Adapter · CIRR MeanR | MeanR: 62.64 -> 53.74 | Hurt |
Closest papers by meaning
- Reason-before-Retrieve: One-Stage Reflective Chain-of-Thoughts for Training-Free Zero-Shot Composed Image Retrieval (CVPR 2025 · similarity 0.91)
- Language-only Training of Zero-shot Composed Image Retrieval (CVPR 2024 · similarity 0.91)
- Context-I2W: Mapping Images to Context-Dependent Words for Accurate Zero-Shot Composed Image Retrieval (AAAI 2024 · similarity 0.90)
- Vision-by-Language for Training-Free Compositional Image Retrieval (ICLR 2024 · similarity 0.90)
- Image2Sentence based Asymmetrical Zero-shot Composed Image Retrieval (ICLR 2024 · similarity 0.90)
- Missing Target-Relevant Information Prediction with World Model for Accurate Zero-Shot Composed Image Retrieval (CVPR 2025 · similarity 0.89)
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.
