List your company
ResearchFormerly WizWand

Mathematical Reasoning on MathVista (Accuracy %)

75.2Accuracy (%)EASE-7B
30.89642.39853.965.402May 18, 2026May 19, 2026May 21, 2026May 23, 2026May 25, 2026May 27, 2026May 29, 2026
Updated 5mo ago

Evaluation results

MethodLinks
EASE-7BModel Scale=7B2026.0575.2
VGPO-7BModel Scale=7B2026.0574.1
ThinkLite-VL-7BModel Scale=7B2026.0573.6
NoisyRollout-7BModel Scale=7B2026.0572.9
PAPO_D-7BModel Scale=7B2026.0572.3
Vision-Zero(CLEVR)2026.0570.6
VPPO-RL-7BModel Scale=7B2026.0570.5
DUELRole=Solver2026.0569.8
EvoLMM2026.0569.7
IVR-R1Backbone model=Qwen3-VL-4B2026.0569.6
MM-UPT2026.0569.4
OpenVLThinker-7B2026.0568.9
Qwen3-VL-32B-InstructBackbone model=Qwen3-VL-32B-Instruct2026.0568.7
DUELRole=Challenger2026.0568.6
Qwen2.5-VL-7B2026.0568.4
Vision-R1Backbone model=Qwen3-VL-4B2026.0568.3
Vision-SR1Backbone model=Qwen3-VL-4B2026.0568.2
VLAA-Thinker-7B2026.0568.2
IVR-R1Backbone model=Qwen2.5-VL-7B2026.0567.9
Vision-R1Backbone model=Qwen2.5-VL-7B2026.0567.4
MM-Eureka-7BModel Scale=7B2026.0566.3
Supervised Fine-tuning (before RL)Backbone model=Qwen3-VL-4B2026.0565.6
VL-Rethinker-7BModel Scale=7B2026.0565.2
Supervised Fine-tuning (before RL)Backbone model=Qwen2.5-VL-7B2026.0563.6
Vision-SR1Backbone model=Qwen2.5-VL-7B2026.0563.5
Zero-shot Inference (before RL)Backbone model=Qwen3-VL-4B2026.0554.5
Qwen2.5-VL-72B-InstructBackbone model=Qwen2.5-VL-72B-Instruct2026.0548.1
Qwen2.5-VL-32B-InstructBackbone model=Qwen2.5-VL-32B-Instruct2026.0537.2
Zero-shot Inference (before RL)Backbone model=Qwen2.5-VL-7B2026.0532.6