List your company
ResearchFormerly WizWand

Mathematical Reasoning on GradeMath (pass@1)

64.76Pass@1SGPO
22.17233.228544.28555.3415May 16, 2025
Updated 8mo ago

Evaluation results

MethodLinks
SGPOBackbone=QwQ-32B, Judge Model=DeepSeek-V3-03242025.0564.76
GRPOBackbone=QwQ-32B2025.0563.81
BaselineBackbone=QwQ-32B2025.0562.38
BaselineBackbone=Qwen2.5-32B-Instruct2025.0553.81
SGPOBackbone=Qwen2.5-32B-Instruct, Judge Model=o4-mini-04162025.0553.81
SGPOBackbone=Qwen2.5-14B-Instruct, Judge Model=o4-mini-04162025.0553.33
GRPOBackbone=Qwen2.5-32B-Instruct2025.0552.86
BaselineBackbone=Qwen2.5-14B-Instruct2025.0549.52
SGPOBackbone=DeepSeek-R1-Distill-Qwen-7B, Judge Model=DeepSeek-V3-03242025.0547.17
GRPOBackbone=Qwen2.5-14B-Instruct2025.0547.14
SGPOBackbone=DeepSeek-R1-Distill-Qwen-7B, Judge Model=Qwen3-235B-A22B2025.0546.67
SGPOBackbone=DeepSeek-R1-Distill-Qwen-7B, Judge Model=o4-mini-04162025.0546.19
SGPOBackbone=DeepSeek-R1-Distill-Qwen-7B, Judge Model=QwQ-32B2025.0545.24
GRPOBackbone=DeepSeek-R1-Distill-Qwen-7B2025.0543.33
SGPOBackbone=DeepSeek-R1-Distill-Qwen-7B, Judge Model=QwQ-32B, Stability Parameters=w/o {beta,gamma}2025.0542.38
BaselineBackbone=DeepSeek-R1-Distill-Qwen-7B2025.0541.43
SGPOBackbone=DeepSeek-R1-Distill-Llama-8B, Judge Model=Claude-3.72025.0529.05
GRPOBackbone=DeepSeek-R1-Distill-Llama-8B2025.0528.33
BaselineBackbone=DeepSeek-R1-Distill-Llama-8B2025.0523.81