Research map / Vision-language and multimodal / CLIP and zero-shot
Manipulation Intention Understanding for Zero-Shot Composed Image Retrieval
Yuanmin Tang, Jing Yu, Keke Gai, Gang Xiong, Gaopeng Gou, Meikang Qiu, Qi Wu
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 intent-CC3M pseudo-manipulation texts to training (ablation) | De-MINDS · CIRR | Recall@1: 26.5 -> 30.0 | Improved |
| add chain-of-thought prompting for text generation (ablation) | De-MINDS · CIRR | Recall@1: 26.7 -> 30.0 | Improved |
| add similarity-guided filtering of generated texts (ablation) | De-MINDS · CIRR | Recall@1: 27.3 -> 30.0 | Improved |
| add the intention-understanding module (learnable queries) (ablation) | De-MINDS · CIRR | Recall@1: 25.9 -> 30.0 | Improved |
| add global feature to the query (ablation) | De-MINDS · CIRR | Recall@1: 27.7 -> 30.0 | Improved |
| add adaptive gating instead of direct summation (ablation) | De-MINDS · CIRR | Recall@1: 28.2 -> 30.0 | Improved |
| add distillation loss (ablation) | De-MINDS · CIRR | Recall@1: 26.7 -> 30.0 | Improved |
| use symmetric contrastive distillation instead of cosine distillation (ablation) | De-MINDS · CIRR | Recall@1: 28.4 -> 30.0 | Improved |
| use LLaVA+GPT-4o texts instead of Qwen-VL generated texts (ablation) | De-MINDS · CIRR | Recall@1: 28.8 -> 30.0 | Improved |
| use Long-CLIP instead of intention understanding (ablation) | De-MINDS · CIRR | Recall@1: 30.0 -> 26.3 | Hurt |
| train baseline Context-I2W with intent-CC3M captions | Context-I2W · CIRR | Recall@1: 25.6 -> 26.5 | Improved |
| use De-MINDS instead of PrediCIR | PrediCIR comparison · CIRR | Recall@1: 27.2 -> 30.0 | 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.
- Context-I2W* — Context-I2W: Mapping Images to Context-Dependent Words for Accurate Zero-Shot Composed Image Retrieval (AAAI 2024) · better on 4 of 4 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
- Context-I2W: Mapping Images to Context-Dependent Words for Accurate Zero-Shot Composed Image Retrieval (AAAI 2024 · similarity 0.88)
- Missing Target-Relevant Information Prediction with World Model for Accurate Zero-Shot Composed Image Retrieval (CVPR 2025 · similarity 0.87)
- Reason-before-Retrieve: One-Stage Reflective Chain-of-Thoughts for Training-Free Zero-Shot Composed Image Retrieval (CVPR 2025 · similarity 0.87)
- Leveraging Large Vision-Language Model as User Intent-Aware Encoder for Composed Image Retrieval (AAAI 2025 · similarity 0.87)
- Unleashing In-context Learning of Autoregressive Models for Few-shot Image Manipulation (CVPR 2025 · similarity 0.86)
- ForgerySleuth: Empowering Multimodal Large Language Models for Image Manipulation Detection (NeurIPS 2025 · 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.
