AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

GEAR-Seg: A Grounded Explainable Agent for Reasoning Segmentation and Data Engine

Yanan Wang, Wen li, Yibin Ying, Zhenghao Fei

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Chain-of-thought multimodal reasoning · 0 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 DAM dense descriptions instead of LLaMA captions (ablation)GEAR-Seg · ReasonSeg valgIoU: 35.3 -> 57.4Improved
add DAM dense descriptions instead of LLaMA captions (ablation)GEAR-Seg · LLM-Seg40k valgIoU: 30.4 -> 52.2Improved
swap DeepSeek-R1 for Gemini-3.1-Pro as reasoning coreGEAR-Seg · ReasonSeg valgIoU: 57.4 -> 58.2Improved
swap DeepSeek-R1 for Gemini-3.1-Pro as reasoning coreGEAR-Seg · LLM-Seg40k valgIoU: 52.2 -> 48.5Hurt
swap DeepSeek-R1 for Qwen3-32BGEAR-Seg · ReasonSeg valgIoU: 57.4 -> 53.9Hurt
swap DeepSeek-R1 for local Qwen 8BGEAR-Seg · StrawDImAP50:95: 0.507 -> 0.512No clear effect
swap DeepSeek-R1 for local Qwen 8BGEAR-Seg · Mega_PeachmAP50:95: 0.368 -> 0.345Hurt
swap DeepSeek-R1 for Sentence-TransformersGEAR-Seg · StrawDImAP50:95: 0.507 -> 0.456Hurt
add explanation text loss (L_txt) to mask loss (ablation)LISA-13B on GEAR-131K · GEAR-131K testgIoU: 26.6 -> 37.2Improved
train on 1.2k GEAR-Seg images instead of 239LISA-13B distilled · ReasonSeg valgIoU: 51.4 -> 57.2Improved
use GEAR-Seg pseudo-labels instead of human labels (239 images)LISA-13B distilled · ReasonSeg valgIoU: 55.6 -> 51.4Hurt
use GEAR-Seg zero-shot instead of fine-tuned LISA-13BGEAR-Seg · ReasonSeg valcIoU: 62.9 -> 47.5Hurt

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