List your company
ResearchFormerly WizWand

Mathematical Reasoning on AMO-Bench (Acc, Tok)

20.5Accuracy (Acc)HINT TUNING
-0.824.71510.2515.785May 9, 2026
Updated 5mo ago

Evaluation results

MethodLinks
HINT TUNINGModel Scale=Qwen3-4B-Thinking-2507, # Data=1K2026.0520.525,926
OriginalModel Scale=DeepSeek-R1-Distill-Qwen-32B, # Data=–2026.0515.421,151
SFT-S1.1Model Scale=DeepSeek-R1-Distill-Qwen-32B, # Data=1K2026.0515.418,836
Budget PromptModel Scale=Qwen3-4B-Thinking-2507, # Data=–2026.0512.825,509
Budget PromptModel Scale=DeepSeek-R1-Distill-Qwen-14B, # Data=–2026.0512.818,541
HINT TUNINGModel Scale=DeepSeek-R1-Distill-Qwen-14B, # Data=1K2026.0512.815,939
HINT TUNINGModel Scale=DeepSeek-R1-Distill-Qwen-32B, # Data=1K2026.0512.818,650
OriginalModel Scale=Qwen3-4B-Thinking-2507, # Data=–2026.0510.327,533
SFT-S1.1Model Scale=DeepSeek-R1-Distill-Qwen-14B, # Data=1K2026.0510.320,041
SFT-S1.1Model Scale=Qwen3-4B-Thinking-2507, # Data=1K2026.057.718,881
SFT-S1.1Model Scale=DeepSeek-R1-Distill-Qwen-7B, # Data=1K2026.057.719,455
OriginalModel Scale=DeepSeek-R1-Distill-Qwen-14B, # Data=–2026.057.719,868
Budget PromptModel Scale=DeepSeek-R1-Distill-Qwen-32B, # Data=–2026.057.717,026
Budget PromptModel Scale=DeepSeek-R1-Distill-Qwen-7B, # Data=–2026.055.120,184
VeriThinkerModel Scale=DeepSeek-R1-Distill-Qwen-7B, # Data=340K2026.055.115,279
HINT TUNINGModel Scale=DeepSeek-R1-Distill-Qwen-7B, # Data=1K2026.055.114,517
OriginalModel Scale=DeepSeek-R1-Distill-Qwen-7B, # Data=–2026.052.620,957
EfficientRModel Scale=DeepSeek-R1-Distill-Qwen-7B, # Data=3.2K†, Alpha=0.12026.052.618,132
EfficientRModel Scale=DeepSeek-R1-Distill-Qwen-7B, # Data=3.2K†, Alpha=0.42026.05018,718
AdaptThinkModel Scale=DeepSeek-R1-Distill-Qwen-7B, # Data=40K†2026.05017,421