AI Paper Atlas

Research map / Video / Video-language understanding

Sa2VA: Marrying SAM2 With MLLM for Dense Grounded Understanding of Images and Videos

Haobo Yuan, Xiangtai Li, Tao Zhang, Yueyi Sun, Zilong Huang, Shilin Xu, Shunping Ji, Yunhai Tong, Lu Qi, Jiashi Feng, Ming–Hsuan Yang

TPAMI 2026 · Video › Video-language understanding · approach: Video LLMs with fine-grained understanding · 206 citations (OpenAlex)

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
add image QA data to co-training (ablation)Sa2VA-1B · MMBench / MMEaccuracy: 63.4 -> 68.3 MMBench; 1298/359 -> 1381/405 MMEImproved
add image segmentation data to co-training (ablation)Sa2VA-1B · RefCOCO / RefCOCO+ / RefCOCOgcIoU: 20.2 -> 77.4; 20.6 -> 69.9; 23.2 -> 72.3Improved
add video QA data to co-training (ablation)Sa2VA-1B · MMBench-Videoscore: 0.71 -> 1.07Improved
add video segmentation data to co-training (ablation)Sa2VA-1B · MeViS / Ref-DAVIS17J&F: 44.4 -> 50.8 MeViS; 69.0 -> 72.3 Ref-DAVIS17Improved
add 3M Infinity-MM image QA samplesSa2VA-1B · MMBenchaccuracy: 68.3 -> 70.4Improved
add Ref-SAV training dataSa2VA-1B · MeViS (val_u)J&F: 50.8 -> 52.5Improved
fine-tune UniRef++ on Ref-SAV training setUniRef++ · Ref-SAV valJ&F: 10.5 -> 14.6Improved
sample uniform 5 frames instead of first 5 frames (ablation)Sa2VA-8B · MeViS (val_u)J&F: 58.9 -> 62.9Improved
use more consecutive first frames (1 to 4) (ablation)Sa2VA-8B · MeViS (val_u)J&F: 55.1 -> 59.5Improved
replace SAM-2 with SAM-3 segmenter (ablation)Sa2VA (Qwen3-VL-4B) · MeViS / ReVOSJ&F: 61.4 -> 65.3 MeViS; 61.4 -> 66.3 ReVOSImproved
use Sa2VA paradigm with LLaVA-1.5-7B instead of LISA-7BLISA-7B · RefCOCO / RefCOCO+ / RefCOCOgcIoU: 74.9 -> 80.3 RefCOCO; 65.1 -> 73.1 RefCOCO+; 67.9 -> 79.2 RefCOCOgImproved
scale Sa2VA from 8B to 26BSa2VA · MeViSJ&F: 46.9 -> 46.2No clear effect

Later papers that compared with this method

5 later accepted papers report Sa2VA-8B* as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.

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.

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