AI Paper Atlas

Research map / Video / Video segmentation and tracking

Towards Streaming Referring Video Segmentation via Large Language Model

Wenkang Zhang, Kaicheng Yang, Xiang An, Qiang Li, Ziyong Feng, Wankou Yang, Jiankang Deng

CVPR 2026 · Video › Video segmentation and tracking · approach: Video semantic and panoptic segmentation

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 changedWhereMeasuredEffect
recycle [SEG] embedding as [INFO] token across frames (SER) (ablation)StreamingRVOS-1B · MeViS (valu)J&F: 52.4 -> 58.5Improved
add SER to streaming Sa2VA baseline (ablation)Sa2VA-1B-Stream · MeViS (valu)J&F: 57.1 -> 58.5Improved
add OMCP video fine-tuning on top of SER (ablation)StreamingRVOS-1B · MeViS (valu)J&F: 58.5 -> 59.5Improved
add OMCP video fine-tuning on top of SER (ablation)StreamingRVOS-1B · Ref-DAVISJ&F: 75.0 -> 76.4Improved
add OMCP video fine-tuning on top of SER (ablation)StreamingRVOS-4B · ReVOS (Referring)J&F: 61.2 -> 63.0Improved
add OMCP video fine-tuning on top of SER (ablation)StreamingRVOS-1B · RefCOCOcIoU: 80.7 -> 80.3Hurt
add SER to Sa2VA streaming baseline (ablation)Sa2VA-1B-Stream · Ref-DAVISJ&F: 74.4 -> 75.0Mixed
store 2 or 3 [INFO] embeddings instead of 1 (ablation)StreamingRVOS-1B · MeViS (valu)J&F: 58.5 -> 58.1 (2), 58.2 (3)No clear effect
store 2 [INFO] embeddings instead of 1 (ablation)StreamingRVOS-1B · ReVOS (Referring)J&F: 58.9 -> 60.2Improved
update [INFO] periodically or via OMCP instead of first frame only (ablation)StreamingRVOS-1B · MeViS (valu)J&F: 54.4 -> 59.5Improved
use OMCP adaptive update instead of updating every frame (ablation)StreamingRVOS-1B · MeViS (valu)J&F: 59.2 -> 59.5Improved
use OMCP adaptive update instead of updating every frame (ablation)StreamingRVOS-1B · speedFPS: 3.4 -> 6.7Improved

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.

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

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy