List your company
ResearchFormerly WizWand

Mathematical Reasoning on AIME 25 (Average@16, Avg.)

49.79Average@16 ScoreTeacher
-1.679611.682725.04538.4073Oct 1, 2025Nov 9, 2025Dec 19, 2025Jan 28, 2026Mar 8, 2026Apr 17, 2026May 27, 2026
Updated 5mo ago

Evaluation results

MethodLinks
TeacherBackbone=QwQ-32B, Prompting=zero-shot2026.0449.7973.9
IRDSModel=QWEN3-4B, Data Budget=10%2026.0526—
IRDSModel=QWEN3-1.7B, Data Budget=10%2026.0516.4—
+CurES-GRPOBase Model=Qwen2.5-Math-7B, Training Method=CurES-GRPO2025.1015.2152.41
+GVM-GRPOBase Model=Qwen2.5-Math-7B, Training Method=GVM-GRPO2025.101550.64
Dr.GRPOBackbone=Llama3.2-3B-Instruct, Domain Type=In-domain2026.0513.1—
+CurES-RPPBase Model=Qwen2.5-Math-7B, Training Method=CurES-RPP2025.1012.9249.62
GRPOBackbone=Llama3.2-3B-Instruct, Domain Type=In-domain2026.0512.5—
+Speed-RL-RPPBase Model=Qwen2.5-Math-7B, Training Method=Speed-RL-RPP2025.1012.0849.56
+RPPBase Model=Qwen2.5-Math-7B, Training Method=RPP2025.1011.6750.18
+CurES-RPPBase Model=Qwen2.5-Math-1.5B, Training Method=CurES-RPP2025.1011.4644.14
+Speed-RL-GRPOBase Model=Qwen2.5-Math-7B, Training Method=Speed-RL-GRPO2025.1011.4648.84
+GVM-GRPOBase Model=Qwen2.5-Math-1.5B, Training Method=GVM-GRPO2025.1011.2542.82
PRIMEBackbone=Llama3.2-3B-Instruct, Domain Type=In-domain2026.0511.2—
+CurES-GRPOBase Model=Qwen2.5-Math-1.5B, Training Method=CurES-GRPO2025.1010.4244.94
+Speed-RL-RPPBase Model=Qwen2.5-Math-1.5B, Training Method=Speed-RL-RPP2025.1010.0841.75
+GRPOBase Model=Qwen2.5-Math-7B, Training Method=GRPO2025.101047.59
Gen-SSDBackbone=Qwen2.5-Math-1.5B, Prompting=zero-shot, Chunk size=4K2026.041034.94
+GVM-RPPBase Model=Qwen2.5-Math-7B, Training Method=GVM-RPP2025.108.3346.95
MoRSDBackbone=Qwen2.5-Math-1.5B, Prompting=zero-shot2026.047.533.07
Standard KDBackbone=Qwen2.5-Math-1.5B, Prompting=zero-shot2026.047.0829.01
+Speed-RL-GRPOBase Model=Qwen2.5-Math-1.5B, Training Method=Speed-RL-GRPO2025.106.8841.34
+GRPOBase Model=Qwen2.5-Math-1.5B, Training Method=GRPO2025.106.6741.64
+GVM-RPPBase Model=Qwen2.5-Math-1.5B, Training Method=GVM-RPP2025.106.0442.21
MCC-KDBackbone=Qwen2.5-Math-1.5B, Prompting=zero-shot2026.045.2130.25
Cold StartBackbone=Qwen2.5-Math-1.5B, Prompting=zero-shot2026.043.7518.3
+RPPBase Model=Qwen2.5-Math-1.5B, Training Method=RPP2025.103.3335.86
Self-DistillationBackbone=Qwen2.5-Math-1.5B, Prompting=zero-shot2026.042.7124.25
Qwen2.5-Math-7BBase Model=Qwen2.5-Math-7B, Training Method=Base2025.101.8833.48
Qwen2.5-Math-1.5BBase Model=Qwen2.5-Math-1.5B, Training Method=Base2025.101.6720
StudentBackbone=Qwen2.5-Math-1.5B, Prompting=two-shot2026.040.8315.54
IRDSModel=LLAMA-3.1-8B-INSTRUCT, Data Budget=10%2026.050.4—
OriginalBackbone=Llama3.2-3B-Instruct, Domain Type=In-domain2026.050.3—