List your company
ResearchFormerly WizWand

Multimodal Mathematical Reasoning on WeMath (test)

72.15AccuracyGSPO + KAWHI (Qwen3)
53.606858.420963.23568.0491Mar 28, 2026
Updated 7mo ago

Evaluation results

MethodLinks
GSPO + KAWHI (Qwen3)Backbone Model=Qwen3-VL-4B-Instruct, Reward Strategy=uniform reward, RL Method=KAWHI2026.0372.1555.54
GSPO (Qwen3)Backbone Model=Qwen3-VL-4B-Instruct, Reward Strategy=uniform reward, RL Method=GSPO2026.0371.5554.14
DAPO + KAWHI (Qwen3)Backbone Model=Qwen3-VL-4B-Instruct, Reward Strategy=uniform reward, RL Method=KAWHI2026.0371.2353.06
GRPO + KAWHI (Qwen3)Backbone Model=Qwen3-VL-4B-Instruct, Reward Strategy=uniform reward, RL Method=KAWHI2026.0370.9851.98
GRPO (Qwen3)Backbone Model=Qwen3-VL-4B-Instruct, Reward Strategy=uniform reward, RL Method=GRPO2026.0370.3151.17
DAPO (Qwen3)Backbone Model=Qwen3-VL-4B-Instruct, Reward Strategy=uniform reward, RL Method=DAPO2026.0369.6751.8
Qwen3-VL-4B-Instruct BaseBackbone Model=Qwen3-VL-4B-Instruct, Reward Strategy=none, RL Method=Base2026.0369.4350.53
GSPO + KAWHIBackbone Model=Qwen2.5-VL-7B-Instruct, Reward Strategy=uniform reward, RL Method=KAWHI2026.0359.7152.94
GSPOBackbone Model=Qwen2.5-VL-7B-Instruct, Reward Strategy=uniform reward, RL Method=GSPO2026.0359.0251.4
DAPO + KAWHIBackbone Model=Qwen2.5-VL-7B-Instruct, Reward Strategy=uniform reward, RL Method=KAWHI2026.0358.2151.85
GRPOBackbone Model=Qwen2.5-VL-7B-Instruct, Reward Strategy=uniform reward, RL Method=GRPO2026.0358.150.18
DAPOBackbone Model=Qwen2.5-VL-7B-Instruct, Reward Strategy=uniform reward, RL Method=DAPO2026.0357.9850.79
GRPO + KAWHIBackbone Model=Qwen2.5-VL-7B-Instruct, Reward Strategy=uniform reward, RL Method=KAWHI2026.0357.8451
VPPOBackbone Model=Qwen2.5-VL-7B-Instruct, Reward Strategy=fine-grained reward, RL Method=VPPO2026.0357.1250.29
FT-RLBackbone Model=Qwen2.5-VL-7B-Instruct, Reward Strategy=fine-grained reward, RL Method=FT-RL2026.0356.7849.37
Step-GRPOBackbone Model=Qwen2.5-VL-7B-Instruct, Reward Strategy=fine-grained reward, RL Method=Step-GRPO2026.0355.3247.82
Qwen2.5-VL-7B-Instruct BaseBackbone Model=Qwen2.5-VL-7B-Instruct, Reward Strategy=none, RL Method=Base2026.0354.3247.68