Research map / Vision-language and multimodal / Human and embodied modeling
Interactive Person Retrieval via Multi-Turn Multimodal Conversation
Yang Bai, Tingfeng Wang, Bin Yang, Min Cao, Jinqiao Wang, Mang Ye
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 ATE (per-turn encoding) in place of concatenating the dialogue: accuracy (ablation) | MNEMO · MInterPEDES MInterPR | R-1: 91.51 -> 88.90 | Hurt |
| add ATE in place of concatenating the dialogue: training time (ablation) | MNEMO · MInterPEDES MInterPR | training time (ms): 661.78 -> 176.25 | Improved |
| add ATE in place of concatenating the dialogue: inference time (ablation) | MNEMO · MInterPEDES MInterPR | inference time (ms): 63.76 -> 22.77 | Improved |
| add DMA on top of ATE (ablation) | MNEMO · MInterPEDES MInterPR | R-1: 88.90 -> 92.26 | Improved |
| add DMA on top of ATE (ablation) | MNEMO · MInterPEDES MInterPR | mAP: 76.99 -> 80.31 | Improved |
| add DMA on top of ATE: extra inference cost (ablation) | MNEMO · MInterPEDES MInterPR | inference time (ms): 22.77 -> 24.07 | No clear effect |
| full MNEMO (ATE+DMA) vs concatenation baseline (ablation) | MNEMO · MInterPEDES MInterPR | R-1: 91.51 -> 92.26 | Improved |
| increase static memory queries from 0 to 16 (ablation) | DMA · MInterPEDES MInterPR | R-1 | Improved |
| increase static memory queries beyond 16 (ablation) | DMA · MInterPEDES MInterPR | R-1 | No clear effect |
| remove dynamic memory queries (ablation) | DMA · MInterPEDES MInterPR | R-1 | Hurt |
| increase dynamic memory queries beyond 16 (ablation) | DMA · MInterPEDES MInterPR | R-1 | Hurt |
| add visual feedback turns on top of question-answering dialogue | MNEMO vs MInterPR scenario · MInterPEDES | R-1: 78.09 -> 92.26 | 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.
- LLaVA-ReID — LLaVA-ReID: Selective Multi-image Questioner for Interactive Person Re-Identification (ICML 2025) · 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
- Chat-based Person Retrieval via Dialogue-Refined Cross-Modal Alignment (CVPR 2025 · similarity 0.92)
- F2RVLM: Boosting Fine-grained Fragment Retrieval for Multi-Modal Long-form Dialogue with Vision Language Model (AAAI 2026 · similarity 0.88)
- SeCom: On Memory Construction and Retrieval for Personalized Conversational Agents (ICLR 2025 · similarity 0.86)
- Beyond Global Similarity: Multi-Conditional Retrieval for Fine-Grained Cross-Modal Understanding (CVPR 2026 · similarity 0.86)
- Human-centered Interactive Learning via MLLMs for Text-to-Image Person Re-identification (CVPR 2025 · similarity 0.85)
- LLaVA-ReID: Selective Multi-image Questioner for Interactive Person Re-Identification (ICML 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.
