List your company
ResearchFormerly WizWand

Mathematical Reasoning on LiveMathBench

74.38AccuracySMCS
-0.895218.647438.1957.7326Jul 14, 2025Jul 31, 2025Aug 18, 2025Sep 5, 2025Sep 23, 2025Oct 11, 2025Oct 29, 2025
Updated 5mo ago

Evaluation results

MethodLinks
SMCS2025.0774.38
Self-MoA2025.0760.33
GPT-4.12025.0759.5
QwQ-32B2025.0758.68
GSPO + TACBackbone=DeepSeek-R1-Distill-Qwen-7B, Training steps=30002025.1024
DRGRPOBackbone=DeepSeek-R1-Distill-Qwen-7B, Training steps=30002025.1021
DeepMathBackbone=DeepSeek-R1-Distill-Qwen-7B, Training steps=30002025.1017
VeriThinkerBackbone=DeepSeek-R1-Distill-Qwen-7B, Training steps=30002025.1014
GSPO + TACBackbone=DeepSeek-R1-Distill-Qwen-1.5B, Training steps=30002025.1014
BASELINEBackbone=DeepSeek-R1-Distill-Qwen-7B, Training steps=30002025.1013
GRPO-LEADBackbone=DeepSeek-R1-Distill-Qwen-7B, Training steps=30002025.1013
ExGRPOBackbone=DeepSeek-R1-Distill-Qwen-7B, Training steps=30002025.1013
SkyWorkBackbone=DeepSeek-R1-Distill-Qwen-7B, Training steps=30002025.1012
BASELINEBackbone=DeepSeek-R1-Distill-Qwen-1.5B, Training steps=30002025.1011
ExGRPOBackbone=DeepSeek-R1-Distill-Qwen-1.5B, Training steps=30002025.1010
Open-RS3Backbone=DeepSeek-R1-Distill-Qwen-1.5B, Training steps=30002025.106
DRA-GRPOBackbone=DeepSeek-R1-Distill-Qwen-1.5B, Training steps=30002025.105
STILL-3Backbone=DeepSeek-R1-Distill-Qwen-1.5B, Training steps=30002025.103
Eurus-2Backbone=DeepSeek-R1-Distill-Qwen-7B, Training steps=30002025.102