List your company

Mathematical Reasoning on MATH (acc, len)

76.67Accuracy (MATH)GRPO-O1
12.470829.137945.80562.4721Sep 5, 2025

Evaluation results

MethodLinks
GRPO-O1Backbone=Gemma3-4B2025.0976.67834
GRPOBackbone=Gemma3-4B2025.0976.541,177
GRPO-ERBackbone=Gemma3-4B2025.0975.73979
SA-GRPOBackbone=Gemma3-4B2025.0975.551,130
Gemma3-4BBackbone=Gemma3-4B, Strategy=Base2025.0968.33837
GRPOBackbone=Qwen2-2B-VL2025.0925.25321
GRPO-ERBackbone=Qwen2-2B-VL2025.0923.73212
GRPO-O1Backbone=Qwen2-2B-VL2025.0923.279
SA-GRPOBackbone=Qwen2-2B-VL2025.0921.2284
Qwen2-2B-VLBackbone=Qwen2-2B-VL, Strategy=Base2025.0914.9415.84