Research map / Video / Video-language understanding
FeVOS: Foresight Expression Video Object Segmentation
Kehan Lan, Kaining Ying, Henghui Ding
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 RL (GRPO) stage after CoT-SFT (ablation) | FeVOS-R1 · FeVOS val | J&F: 37.2 -> 42.3 | Improved |
| add CoT-SFT stage before RL (ablation) | FeVOS-R1 · FeVOS val | J&F: 36.0 -> 42.3 | Improved |
| CoT-SFT only vs RL only from scratch (ablation) | FeVOS-R1 · FeVOS val | J&F: 36.0 -> 37.2 | Improved |
| add format reward to IoU reward (ablation) | FeVOS-R1 GRPO · FeVOS val | J&F: 42.3 -> 40.9 | Hurt |
| use IoU reward instead of format reward only (ablation) | FeVOS-R1 GRPO · FeVOS val | J&F: 37.7 -> 42.3 | Improved |
| fine-tune Sa2VA on FeVOS with standard SFT | Sa2VA · FeVOS val | J&F: 25.4 -> 35.8 | Improved |
| fine-tune GLUS on FeVOS | GLUS · FeVOS val | J&F: 29.6 -> 33.5 | Improved |
| CoT-SFT plus GRPO training vs standard SFT baseline | Sa2VA* · FeVOS val | J&F: 35.8 -> 42.3 | Improved |
| fine-tune Sa2VA on FeVOS (standard SFT) | Sa2VA · ReVOS | J&F: 59.1 -> 58.1 | Hurt |
| CoT-SFT plus GRPO training on FeVOS | Sa2VA* · ReVOS reasoning subset | J&F: 55.2 -> 57.8 | Improved |
| CoT-SFT plus GRPO training on FeVOS | Sa2VA* · MeViS | J&F: 46.5 -> 49.5 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 3 on most shared results.
- GLUS — GLUS: Global-Local Reasoning Unified into A Single Large Language Model for Video Segmentation (CVPR 2025) · better on 1 of 2 shared results
- VideoGLaMM — VideoGLaMM : A Large Multimodal Model for Pixel-Level Visual Grounding in Videos (CVPR 2025) · better on 1 of 1 shared result
- 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
- Online Reasoning Video Object Segmentation (ECCV 2026 · similarity 0.89)
- VIRST: Video-Instructed Reasoning Assistant for SpatioTemporal Segmentation (CVPR 2026 · similarity 0.87)
- Training-Free Spatio-temporal Decoupled Reasoning Video Segmentation with Adaptive Object Memory (AAAI 2026 · similarity 0.87)
- TOMATO: Assessing Visual Temporal Reasoning Capabilities in Multimodal Foundation Models (ICLR 2025 · similarity 0.85)
- SciVideoBench: Benchmarking Scientific Video Reasoning in Large Multimodal Models (ICML 2026 · similarity 0.85)
- Look, Remember and Reason: Grounded Reasoning in Videos with Language Models (ICLR 2024 · 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.
