List your company
ResearchFormerly WizWand

Multimodal Reasoning on MathVision (mean@8 acc %)

51.36Mean@8 AccuracyRAPO_D
14.367223.971133.57543.1789May 10, 2026
Updated 5mo ago

Evaluation results

MethodLinks
RAPO_DBase Model=Qwen2-VL-8B Instruct, Temperature=0.62026.0551.36
VPPO_DBase Model=Qwen2-VL-8B Instruct, Temperature=0.62026.0550.99
DAPOBase Model=Qwen2-VL-8B Instruct, Temperature=0.62026.0550.33
PAPO_DBase Model=Qwen2-VL-8B Instruct, Temperature=0.62026.0549.67
RAPOBase Model=Qwen3-VL-8B-Instruct2026.0548.27
GRPOBase Model=Qwen3-VL-8B-Instruct2026.0548.03
VPPOBase Model=Qwen3-VL-8B-Instruct2026.0548.03
PAPOBase Model=Qwen3-VL-8B-Instruct2026.0543.75
BaseBase Model=Qwen3-VL-8B-Instruct2026.0541.74
Base (Qwen2-VL-8B Instruct)Temperature=0.62026.0541.74
VL-RethinkerBase Model=Qwen2.5-VL-7B-Instruct2026.0534.21
RAPOBase Model=Qwen3-VL-2B-Instruct2026.0533.55
MM-EurekaBase Model=Qwen2.5-VL-7B-Instruct2026.0532.57
VPPOBase Model=Qwen3-VL-2B-Instruct2026.0530.59
RAPO_DBase Model=Qwen2.5-VL-7B-Instruct2026.0530.26
GRPOBase Model=Qwen3-VL-2B-Instruct2026.0528.29
PAPOBase Model=Qwen3-VL-2B-Instruct2026.0528.29
RAPO_GBase Model=Qwen2.5-VL-7B-Instruct2026.0527.96
BaseBase Model=Qwen3-VL-2B-Instruct2026.0527.3
R1-ShareVLBase Model=Qwen2.5-VL-7B-Instruct2026.0525.99
RAPO_DBase Model=Qwen2-VL-7B-Instruct2026.0523.36
RAPO_GBase Model=Qwen2-VL-7B-Instruct2026.0522.37
MINT-CoTBase Model=Qwen2-VL-7B-Instruct2026.0522.04
BaseBase Model=Qwen2.5-VL-7B-Instruct2026.0521.05
TVCBase Model=Qwen2-VL-7B-Instruct2026.0518.75
BaseBase Model=Qwen2-VL-7B-Instruct2026.0515.79