List your company

Scientific Reasoning on GPQA-D (% Avg@4)

60.4% Avg@4RePro
58.63259.09159.5560.009Dec 1, 2025

Evaluation results

MethodLinks
ReProBackbone=Qwen3-8B, Base Training=GRPO2025.1260.4
GRPOBackbone=Qwen3-8B2025.1259.5
OriginalBackbone=Qwen3-8B2025.1258.7