Research map / Video / Video-language understanding
VISA: Reasoning Video Object Segmentation via Large Language Model
Cilin Yan, Haochen Wang, Shilin Yan, Xiaolong Jiang, Yao Hu, Guoliang Kang, Weidi Xie, Efstratios Gavves
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 |
|---|---|---|---|
| remove image datasets from training (ablation) | VISA · ReVOS val (reasoning) | J&F: 43.0 -> 33.5 | Hurt |
| add Referring VOS datasets to training (ablation) | VISA · ReVOS val (reasoning) | J&F: 39.9 -> 43.0 | Improved |
| add ReVOS instruction tuning data (ablation) | VISA · ReVOS val (reasoning) | J&F: 39.2 -> 43.0 | Improved |
| add video VQA data to training (ablation) | VISA · ReVOS val (reasoning) | J&F: 41.4 -> 43.0 | Improved |
| add ReVOS instruction tuning data on referring segmentation (ablation) | VISA · ReVOS val (referring) | J&F: 52.9 -> 50.9 | No clear effect |
| use Text-guided Frame Sampler target frame instead of first frame (ablation) | VISA · ReVOS overall | J&F: 42.6 -> 44.3 | Improved |
| add 12 reference frames with Global-Local sampling (ablation) | VISA · ReVOS overall | J&F: 44.3 -> 46.9 | Improved |
| use Global-Local sampling instead of Global or Local only (ablation) | VISA · ReVOS overall | J&F: 46.7 -> 46.9 | No clear effect |
| reduce visual tokens per frame from 112 to 56 (ablation) | VISA · ReVOS (reasoning) | J&F: 43.0 -> 39.0 | Hurt |
| increase visual tokens per frame from 112 to 256 (ablation) | VISA · ReVOS (reasoning) | J&F: 43.0 -> 43.2 | No clear effect |
| chain video VQA (LLaMA-VID) with referring VOS (LMPM) for reasoning queries | LMPM · ReVOS (reasoning) | J&F: 18.8 -> 18.2 | Hurt |
| VISA versus LISA on refCOCO val image segmentation | VISA · refCOCO val | cIoU: 74.9 -> 72.4 | Hurt |
Later papers that compared with this method
23 later accepted papers report VISA-13B as a baseline in their results tables; 23 of them report a better number on most of the results they share with it.
- SPOT: Spatiotemporal Prompt Optimization for Motion-Stabilized MLLM-Guided Video Segmentation (CVPR 2026 · better on 5 of 5 shared results)
- Training-Free Spatio-temporal Decoupled Reasoning Video Segmentation with Adaptive Object Memory (AAAI 2026 · better on 4 of 4 shared results)
- Sa2VA: Marrying SAM2 With MLLM for Dense Grounded Understanding of Images and Videos (TPAMI 2026 · better on 3 of 3 shared results)
- CoT-RVS: Zero-Shot Chain-of-Thought Reasoning Segmentation for Videos (ICLR 2026 · better on 2 of 2 shared results)
- Deforming Videos to Masks: Flow Matching for Referring Video Segmentation (ICLR 2026 · better on 2 of 2 shared results)
- PixFoundation 2.0: Do Video Multi-Modal LLMs Use Motion in Visual Grounding? (NeurIPS 2026 · better on 2 of 2 shared results)
- Where It Moves, It Matters: Referring Surgical Instrument Segmentation via Motion (AAAI 2026 · better on 2 of 2 shared results)
- Online Reasoning Video Object Segmentation (ECCV 2026 · better on 2 of 2 shared results)
15 more not listed.
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- OnlineRefer — OnlineRefer: A Simple Online Baseline for Referring Video Object Segmentation (ICCV 2023) · better on 0 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
- One Token to Seg Them All: Language Instructed Reasoning Segmentation in Videos (NeurIPS 2024 · similarity 0.91)
- LISA: Reasoning Segmentation via Large Language Model (CVPR 2024 · similarity 0.89)
- The Devil is in Temporal Token: High Quality Video Reasoning Segmentation (CVPR 2025 · similarity 0.89)
- Fast Reasoning Segmentation for Images and Videos (CVPR 2026 · similarity 0.88)
- MMR: A Large-scale Benchmark Dataset for Multi-target and Multi-granularity Reasoning Segmentation (ICLR 2025 · similarity 0.86)
- DR$^2$Seg: Decomposed Two-Stage Rollouts for Efficient Reasoning Segmentation in Multimodal Large Language Models (ICML 2026 · 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.
