AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Visual-RFT: Visual Reinforcement Fine-Tuning

Ziyu Liu, Zeyi Sun, Yuhang Zang, Xiaoyi Dong, Yuhang Cao, Haodong Duan, Dahua Lin, Jiaqi Wang

ICCV 2025 · Vision-language and multimodal › Multimodal LLMs · approach: Reinforcement learning for visual reasoning · 512 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
replace SFT with Visual-RFT (GRPO with verifiable rewards)Qwen2-VL-2B · Fine-grained classification one-shotaccuracy: +24.3Improved
use SFT on one-shot classificationQwen2-VL-2B · Fine-grained classification one-shotaccuracy: -4.3Hurt
replace SFT with Visual-RFTQwen2-VL-2B · Fine-grained classification 16-shotaccuracy: 64.0 -> 85.3Improved
replace SFT with Visual-RFTQwen2-VL-2B · COCO 8 categories 2-shotmAP: 21.0 -> 41.5Improved
replace SFT with Visual-RFTQwen2-VL-2B · COCO 8 categories 16-shotmAP: 31.3 -> 46.8Improved
replace SFT with Visual-RFTQwen2-VL-7B · COCO 8 categories 4-shotmAP: 44.1 -> 54.3Improved
replace SFT with Visual-RFTQwen2-VL-2B · LVIS 6 rare categories 10-shotmAP: 10.0 -> 19.4Improved
replace SFT with Visual-RFTQwen2-VL-2B · MG 5 categories 4-shot (out-of-domain)mAP: 26.8 -> 61.8Improved
replace SFT with Visual-RFTQwen2-VL-2B · LISA reasoning groundingmIoU test: 28.3 -> 37.6Improved
replace SFT with Visual-RFTQwen2-VL-7B · LISA reasoning groundingmIoU test: 39.1 -> 43.9Improved
use SFT on 7B for COCO novel classesQwen2-VL-7B · COCO open-vocab novelmAP novel: 26.3 -> 25.7No clear effect
replace SFT with Visual-RFTQwen2-VL-7B · COCO open-vocab novelmAP novel: 25.7 -> 35.8Improved

Later papers that compared with this method

4 later accepted papers report Visual-RFT as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.

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