AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

StAR: Segment Anything Reasoner

Seokju Yun, Dongheon Lee, Noori Bae, Jaesung Jun, Chanseul Cho, Youngmin Ro

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Chain-of-thought multimodal reasoning · 4 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
LoRA (rank 16, lr 1e-5, no KL) then rank 64 instead of full tuningVisionReasoner-7B · ReasonSeg-X test / ReasonSeg-RgIoU: 42.2 -> 45.4 / 64.8 -> 66.1Improved
Add SAM-level tiered mask IoU reward and scale rollouts 8 to 16StAR stage-1 · ReasonSeg-X test / ReasonSeg-RgIoU: 45.4 -> 47.1 / 66.1 -> 66.6Improved
Stage-2 training with REST on ReasonSeg-X trainStAR-7B · ReasonSeg-X / ReasonSeg-RgIoU: +1.5 / +1.6Improved
Add label prediction before coordinates (ablation)StAR-7B · ReasonSeg-X / ReasonSeg-RgIoU: 48.6/68.2 -> 49.2/69.7Improved
Predict label last instead of first (ablation)StAR-7B · ReasonSeg-X testgIoU: 49.2 -> 46.3Hurt
Predict label last instead of first (ablation)StAR-8B · ReasonSeg-RcIoU: 69.2 -> 70.8 cIoU; gIoU 73.8 -> 73.1Mixed
Increase REST rollouts n from 16 to 256 (ablation)StAR-8B + MV · ReasonSeg-X testgIoU: 58.1 -> 59.6Improved
Increase REST rollouts n from 16 to 256 (ablation)StAR-7B (no MV) · ReasonSeg-X testgIoU: 48.4 -> 49.2 (peak 49.4 at n=128)No clear effect
Add mask-level majority voting (32 samples)StAR-7B · ReasonSeg-X testgIoU: 49.2 -> 50.3Improved
Add mask-level majority votingStAR-Qwen3-VL 32B · ReasonSeg-X testgIoU: 61.7 -> 64.1Improved
Add mask-level majority votingStAR-Qwen2.5-VL 7B · ReasonSeg-X C/MHgIoU: 28.0 -> 27.9No clear effect
Scale base model from 8B to 32BStAR · ReasonSeg-X testgIoU: 57.9 -> 61.7Improved

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