List your company

Mathematical Reasoning on Olympiad (Accuracy @avg1)

0.457Accuracy (@avg1)LESS
0.230280.289140.3480.40686Nov 30, 2025

Evaluation results

MethodLinks
LESSBackbone=Qwen2.5-Math-7B2025.110.457
GRPOBackbone=Qwen2.5-Math-7B2025.110.449
KL-CovBackbone=Qwen2.5-Math-7B2025.110.437
Forking TokensBackbone=Qwen2.5-Math-7B2025.110.422
GRPOBackbone=Qwen2.5-Base-7B2025.110.409
KL-CovBackbone=Qwen2.5-Base-7B2025.110.408
LESSBackbone=Qwen2.5-Base-7B2025.110.408
Forking TokensBackbone=Qwen2.5-Base-7B2025.110.406
GRPOBackbone=Qwen2.5-Math-1.5B2025.110.396
LESSBackbone=Qwen2.5-Math-1.5B2025.110.396
Forking TokensBackbone=Qwen2.5-Math-1.5B2025.110.385
KL-CovBackbone=Qwen2.5-Math-1.5B2025.110.384
Base LLMBackbone=Qwen2.5-Math-7B2025.110.38
Base LLMBackbone=Qwen2.5-Math-1.5B2025.110.265
Base LLMBackbone=Qwen2.5-Base-7B2025.110.239