AI Paper Atlas

Research map / Detection and segmentation / CNN features

Better, Stronger, Faster: Tackling the Trilemma in MLLM-based Segmentation with Simultaneous Textual Mask Prediction

Jiazhen Liu, Mingkuan Feng, Long Chen

CVPR 2026 · Detection and segmentation › CNN features · approach: Context modeling for semantic segmentation · 6 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 hybrid attention (bidirectional over mask tokens) (ablation)STAMP-2B · single-object RES avgcIoU: 76.2 -> 79.1Improved
add visually-enhanced mask embedding (fuse patch features) (ablation)STAMP-2B · single-object RES avgcIoU: 73.3 -> 79.1Improved
add SAM refinement (ablation)STAMP-2B · single-object RES avgcIoU: 75.3 -> 79.1Improved
lower input resolution from 896 to 504 (ablation)STAMP-2B · single-object RES avgcIoU: 79.1 -> 77.1Mixed
mix LLaVA VQA data with segmentation data in training (ablation)STAMP-2B · RefCOCO valcIoU: 81.9 -> 82.2Improved
mix segmentation data into VQA trainingSTAMP-2B · MMBenchaccuracy: 66.5 -> 68.7No clear effect
use embedding-prediction segmentation heads (LISA, READ) with mixed trainingLISA-7B · general VQAaccuracy: 0 on MMBenchHurt
swap backbone from Qwen2-VL to LLaVA Vicuna-7B (ablation)STAMP · single-object RES avgcIoU: 79.1 (2B) -> 77.0 (Vicuna-7B)No clear effect
use STAMP-7B instead of Text4Seg-13B without SAMMLLM segmentation · gRefCOCO avgavg of gIoU/cIoU: 67.8 -> 71.8Improved
scale STAMP from 2B to 7BSTAMP · single-object RES avgcIoU: 79.1 -> 80.7Improved

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