List your company
ResearchFormerly WizWand

Mathematical Reasoning on MINERVA (Pass@1, Avg.)

41.54Pass@1LoRA
10.194418.332226.4734.6078Oct 1, 2025Nov 12, 2025Dec 24, 2025Feb 4, 2026Mar 18, 2026Apr 29, 2026Jun 11, 2026
Updated 4mo ago

Evaluation results

MethodLinks
LoRAModel=7B2026.0641.54—
LoRA-αModel=7B2026.0641.54—
Base modelModel=7B2026.0640.8—
Full Fine-TuningModel=7B2026.0640.8—
LoRA (10× LR)Model=7B2026.0640.07—
+RPPBase Model=Qwen2.5-Math-7B, Training Method=RPP2025.1039.7150.18
+CurES-GRPOBase Model=Qwen2.5-Math-7B, Training Method=CurES-GRPO2025.1037.8752.41
+Speed-RL-RPPBase Model=Qwen2.5-Math-7B, Training Method=Speed-RL-RPP2025.1037.1349.56
+CurES-RPPBase Model=Qwen2.5-Math-7B, Training Method=CurES-RPP2025.1033.0949.62
+GVM-GRPOBase Model=Qwen2.5-Math-7B, Training Method=GVM-GRPO2025.1032.7250.64
+GRPOBase Model=Qwen2.5-Math-7B, Training Method=GRPO2025.1031.9947.59
+CurES-GRPOBase Model=Qwen2.5-Math-1.5B, Training Method=CurES-GRPO2025.1031.6244.94
Qwen2.5-Math-1.5B + MoPPSBackbone=Qwen2.5-Math-1.5B, Training Method=MoPPS2025.1031.62—
Qwen2.5-Math-1.5B + CurES-GRPOBackbone=Qwen2.5-Math-1.5B, Training Method=CurES-GRPO2025.1031.62—
Full Fine-TuningModel=1.5B2026.0631.25—
+Speed-RL-GRPOBase Model=Qwen2.5-Math-7B, Training Method=Speed-RL-GRPO2025.1029.4148.84
LoRA-αModel=1.5B2026.0629.41—
GRPO + PIPOModel=Qwen3-8B-Base2026.0429.4—
+GVM-RPPBase Model=Qwen2.5-Math-7B, Training Method=GVM-RPP2025.1029.0446.95
LoRA (10× LR)Model=1.5B2026.0629.04—
DAPO + PIPOModel=Qwen3-8B-Base2026.0429—
+CurES-RPPBase Model=Qwen2.5-Math-1.5B, Training Method=CurES-RPP2025.1028.3144.14
GRPOModel=Qwen3-8B-Base2026.0428.3—
DAPOModel=Qwen3-8B-Base2026.0428.3—
+GRPOBase Model=Qwen2.5-Math-1.5B, Training Method=GRPO2025.1027.9441.64
+Speed-RL-RPPBase Model=Qwen2.5-Math-1.5B, Training Method=Speed-RL-RPP2025.1027.9441.75
Qwen2.5-Math-1.5B + DAPOBackbone=Qwen2.5-Math-1.5B, Training Method=DAPO2025.1027.94—
GRPO + PIPOModel=Qwen3-4B-Base2026.0427.9—
DAPO + PIPOModel=Qwen3-4B-Base2026.0427.9—
GSPOModel=Qwen3-8B-Base2026.0427.6—
LoRAModel=1.5B2026.0627.57—
+Speed-RL-GRPOBase Model=Qwen2.5-Math-1.5B, Training Method=Speed-RL-GRPO2025.1027.441.34
+GVM-GRPOBase Model=Qwen2.5-Math-1.5B, Training Method=GVM-GRPO2025.1027.2142.82
GSPOModel=Qwen3-4B-Base2026.0427.2—
DAPOModel=Qwen3-4B-Base2026.0427—
GSPO + PIPOModel=Qwen3-8B-Base2026.0426.4—
GRPOModel=Qwen3-4B-Base2026.0425.4—
+GVM-RPPBase Model=Qwen2.5-Math-1.5B, Training Method=GVM-RPP2025.1024.6342.21
GSPO + PIPOModel=Qwen3-4B-Base2026.0424.3—
Qwen2.5-Math-7BBase Model=Qwen2.5-Math-7B, Training Method=Base2025.1022.7933.48
+RPPBase Model=Qwen2.5-Math-1.5B, Training Method=RPP2025.1021.3235.86
Base ModelModel=Qwen3-8B-Base2026.0417.3—
Base ModelModel=Qwen3-4B-Base2026.0414—
Base modelModel=1.5B2026.0613.33—
Qwen2.5-Math-1.5BBase Model=Qwen2.5-Math-1.5B, Training Method=Base2025.1011.420
Qwen2.5-Math-1.5BBackbone=Qwen2.5-Math-1.5B, Training Method=Base2025.1011.4—