Research map / Vision-language and multimodal / Human and embodied modeling
Ask&Confirm: Active Detail Enriching for Cross-Modal Retrieval With Partial Query
Guanyu Cai, Jun Zhang, Xinyang Jiang, Yifei Gong, Lianghua He, Fufu Yu, Pai Peng, Xiaowei Guo, Feiyue Huang, Xing Sun
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 Ask&Confirm interaction (Q1/A10) | S-SCAN · Visual Genome (Q1/A10) | R@10: 20.4 -> 59.8 | Improved |
| add Ask&Confirm interaction (Q2/A5) | S-SCAN · Visual Genome (Q2/A5) | R@10: 41.7 -> 67.7 | Improved |
| add Ask&Confirm interaction (Q4/A3) | S-SCAN · Visual Genome (Q4/A3) | R@10: 65.9 -> 80.1 | Improved |
| add Ask&Confirm interaction (Q4/A3) | S-SCAN · Visual Genome (Q4/A3) | R@1: 33.5 -> 34.1 | No clear effect |
| add Ask&Confirm interaction (Q1/A10) | T-CMPL · Visual Genome (Q1/A10) | R@10: 16.3 -> 37.0 | Improved |
| add Ask&Confirm interaction (Q4/A3) | T-CMPL · Visual Genome (Q4/A3) | R@1: 14.5 -> 15.1 | No clear effect |
| add Ask&Confirm interaction (Q1/A10) | S-SCAN · Visual Genome (Q1/A10) | Mean Rank: 416.0 -> 96.0 | Improved |
| provide ground-truth object descriptions to the query | S-SCAN · Visual Genome (partial query) | R@10: 20.4 -> 78.4 | Improved |
| use learned RL policy instead of QACohe policy (Q1/A10) (ablation) | Ask&Confirm · Visual Genome (Q1/A10) | Mean Rank: 154.4 -> 96.0 | Improved |
| use learned RL policy instead of QACohe policy (Q4/A3) (ablation) | Ask&Confirm · Visual Genome (Q4/A3) | R@1: 33.8 -> 34.1 | No clear effect |
| use learned RL policy instead of the second-best pre-defined policy (ablation) | Ask&Confirm · Visual Genome (Q1/A10, Q2/A5, Q4/A3) | R@10: 12.1%, 7.2%, 5.0% gain | Improved |
| increase number of actions N_A per round (ablation) | Ask&Confirm · Visual Genome | R@10 | Improved |
Closest papers by meaning
- RMIR: A Benchmark Dataset for Reasoning-Intensive Multimodal Image Retrieval (CVPR 2026 · similarity 0.86)
- ARK: A Dual-Axis Multimodal Retrieval Benchmark along Reasoning and Knowledge (NeurIPS 2026 · similarity 0.86)
- Instance-Level Composed Image Retrieval (NeurIPS 2025 · similarity 0.85)
- Reason-before-Retrieve: One-Stage Reflective Chain-of-Thoughts for Training-Free Zero-Shot Composed Image Retrieval (CVPR 2025 · similarity 0.85)
- MagicLens: Self-Supervised Image Retrieval with Open-Ended Instructions (ICML 2024 · similarity 0.85)
- Multimodal Hypothetical Summary for Retrieval-based Multi-image Question Answering (AAAI 2025 · 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.
