AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Syn-GRPO: Self-Evolving Data Synthesis for MLLM Perception Reasoning

Qihan Huang, Haofei Zhang, Rong Wei, Yi Wang, Rui Tang, Mingli Song, Jie Song

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Reinforcement learning for visual reasoning · 2 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 online image synthesis with diversity rewardGRPO · LISA-Grounding (3B)accuracy: 62.24 -> 68.28Improved
add online image synthesis with diversity rewardGRPO · LISA-Grounding (7B)accuracy: 65.26 -> 70.14Improved
add diversity reward to Syn-GRPO (ablation)Syn-GRPO · LISA-Grounding (3B)accuracy: 64.60 -> 68.28Improved
add diversity reward to Syn-GRPO (ablation)Syn-GRPO · RefCOCOg (3B)accuracy: 85.85 -> 87.45Improved
add entropy loss to GRPOGRPO · RefCOCO+ (3B)accuracy: 84.10 -> 82.55Hurt
add entropy loss / entropy advantage to GRPOGRPO · LISA-Grounding (3B)accuracy: 62.24 -> 64.23Improved
replace offline GPT-4o descriptions with online self-evolving synthesisGRPO + Offline Generation · LISA-Grounding (3B)accuracy: 64.23 -> 68.28Improved
train with diversity smoothing (gamma 0.7 vs 0.9) (ablation)Syn-GRPO · LISA-Grounding (7B)accuracy: 69.24 -> 70.14Improved
use asynchronous instead of synchronous data server (ablation)Syn-GRPO · REC training time (3B)hours: 13.16 -> 6.45Improved
use Syn-GRPO instead of GRPOGRPO · OVD COCO_filtermAP: 18.66 -> 23.74Improved
use Syn-GRPO instead of GRPOGRPO · ISR perceptionmAP: 33.46 -> 35.77Improved
increase training data size 400 to 2000 (ablation)Syn-GRPO · REC LISA-GroundingaccuracyImproved

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