Research map / Video / Video-language understanding
Refer-Agent: A Collaborative Multi-Agent System with Reasoning and Reflection for Referring Video Object Segmentation
Haichao Jiang, Tianming Liang, Wei-Shi Zheng, Jian-Fang Hu
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 Existence Reflection (stage 1) (ablation) | Refer-Agent · ReasonVOS | J&F: 66.6 -> 69.8 | Improved |
| add Consistency Reflection (stage 2) (ablation) | Refer-Agent · ReasonVOS | J&F: 65.2 -> 69.8 | Improved |
| add both reflection stages (ablation) | Refer-Agent · ReasonVOS | J&F: 64.5 -> 69.8 | Improved |
| use Dynamic Focus Layout instead of single keyframe (ablation) | Refer-Agent · ReasonVOS | J&F: 67 -> 69.8 | Improved |
| use Dynamic Focus Layout instead of uniform frame integration (ablation) | Refer-Agent · ReasonVOS | J&F: 67.7 -> 69.8 | Improved |
| increase MAX_TURN from 0 to 4 (ablation) | Refer-Agent · ReasonVOS | J&F: 64.5 -> 69.8 | Improved |
| increase MAX_TURN from 4 to 6 (ablation) | Refer-Agent · ReasonVOS | J&F: 69.8 -> 69.3 | No clear effect |
| keep separate agents instead of merging all tasks into one MLLM call (ablation) | Refer-Agent · ReasonVOS | J&F: 62.5 -> 69.8 | Improved |
| swap Qwen2.5-VL-7B for Ovis2.5-9B as MLLM (ablation) | Refer-Agent · ReasonVOS | J&F: 55.5 -> 69.8 | Improved |
| Refer-Agent (zero-shot) vs SFT RGA3 using same Qwen2.5-VL-7B | Refer-Agent · ReasonVOS | J&F: 53.6 -> 55.5 | Improved |
| Refer-Agent vs SFT methods on GroundMoRe | Refer-Agent · GroundMoRe | J&F: 27.2 -> 33.4 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- GLUS — GLUS: Global-Local Reasoning Unified into A Single Large Language Model for Video Segmentation (CVPR 2025) · better on 2 of 2 shared results
- VISA — VISA: Reasoning Video Object Segmentation via Large Language Model (ECCV 2024) · better on 1 of 1 shared result
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
- Temporal Collection and Distribution for Referring Video Object Segmentation (ICCV 2023 · similarity 0.90)
- Think Before You Segment: An Object-aware Reasoning Agent for Referring Audio-Visual Segmentation (AAAI 2026 · similarity 0.90)
- End-to-End Referring Video Object Segmentation With Multimodal Transformers (CVPR 2022 · similarity 0.89)
- OnlineRefer: A Simple Online Baseline for Referring Video Object Segmentation (ICCV 2023 · similarity 0.89)
- Language As Queries for Referring Video Object Segmentation (CVPR 2022 · similarity 0.89)
- LoSh: Long-Short Text Joint Prediction Network for Referring Video Object Segmentation (CVPR 2024 · similarity 0.88)
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.
