List your company
ResearchFormerly WizWand

Mathematical Reasoning on AIME 24 (Pass@1)

68.13Pass@1MulFeRL
-2.68183215.70200934.0858552.469691Nov 2, 2025Nov 17, 2025Dec 2, 2025Dec 17, 2025Jan 1, 2026Jan 16, 2026Feb 1, 2026
Updated 8mo ago

Evaluation results

MethodLinks
MulFeRLBackbone=Qwen3-4B-Inst, Training Strategy=Multi-turn Feedback-guided Reinforcement Learning2026.0168.13
Critique-GRPOBackbone=Qwen3-4B-Inst, Training Strategy=Reinforcement Learning-based Finetuning2026.0162.53
GRPOBackbone=Qwen3-4B-Inst, Training Strategy=Reinforcement Learning-based Finetuning2026.0157.47
Dr.GRPOBackbone=Qwen3-4B-Inst, Training Strategy=Reinforcement Learning-based Finetuning2026.0156.8
MulFeRLBackbone=Qwen2.5-7B-Base, Training Strategy=Multi-turn Feedback-guided Reinforcement Learning2026.0123.73
CITL-FTBackbone=Qwen3-4B-Inst, Training Strategy=Supervised Learning-based Finetuning2026.0123.07
RAFTBackbone=Qwen3-4B-Inst, Training Strategy=Supervised Learning-based Finetuning2026.0122.93
SFTBackbone=Qwen3-4B-Inst, Training Strategy=Supervised Learning-based Finetuning2026.0122.27
Critique-GRPOBackbone=Qwen2.5-7B-Base, Training Strategy=Reinforcement Learning-based Finetuning2026.0120.93
Qwen3-4B-InstBackbone=Qwen3-4B-Inst, Training Strategy=Base Model2026.0119.73
GRPOBackbone=Qwen2.5-7B-Base, Training Strategy=Reinforcement Learning-based Finetuning2026.0116.13
CITL-FTBackbone=Qwen2.5-7B-Base, Training Strategy=Supervised Learning-based Finetuning2026.0115.2
Dr.GRPOBackbone=Qwen2.5-7B-Base, Training Strategy=Reinforcement Learning-based Finetuning2026.0114.67
SetPO+GRPOBackbone=Qwen2.5-Math-1.5B, Inference Temperature=0.5, Sampling Count (k)=642026.0213.4
SetPO+DAPOBackbone=Qwen2.5-Math-1.5B, Inference Temperature=0.5, Sampling Count (k)=642026.0213.4
SetPO+GSPOBackbone=Qwen2.5-Math-1.5B, Inference Temperature=0.5, Sampling Count (k)=642026.0213.2
Qwen2.5-7B-BaseBackbone=Qwen2.5-7B-Base, Training Strategy=Base Model2026.0113.2
DAPOBackbone=Qwen2.5-Math-1.5B, Inference Temperature=0.5, Sampling Count (k)=642026.0212
SFTBackbone=Qwen2.5-7B-Base, Training Strategy=Supervised Learning-based Finetuning2026.0112
GRPOBackbone=Qwen2.5-Math-1.5B, Inference Temperature=0.5, Sampling Count (k)=642026.0211.5
GSPOBackbone=Qwen2.5-Math-1.5B, Inference Temperature=0.5, Sampling Count (k)=642026.0211.1
RAFTBackbone=Qwen2.5-7B-Base, Training Strategy=Supervised Learning-based Finetuning2026.019.33
Qwen2.5-Math-1.5BBackbone=Qwen2.5-Math-1.5B, Inference Temperature=0.5, Sampling Count (k)=642026.024.6
PREPOBase Model=Qwen3-4B, Training Strategy=PREPO, # Rollouts=348K2025.110.8
Random SelectionBase Model=Qwen3-4B, Training Strategy=Random Selection, # Rollouts=553K2025.110.7
GRESOBase Model=Qwen3-4B, Training Strategy=GRESO, # Rollouts=472K2025.110.6917
Qwen3-4BBase Model=Qwen3-4B, Training Strategy=Base, # Rollouts=–2025.110.5333
Random SelectionBase Model=Qwen2.5-Math-7B, Training Strategy=Random Selection, # Rollouts=905K2025.110.2667
PREPOBase Model=Qwen2.5-Math-7B, Training Strategy=PREPO, # Rollouts=540K2025.110.2615
GRESOBase Model=Qwen2.5-Math-7B, Training Strategy=GRESO, # Rollouts=654K2025.110.2583
Qwen2.5-Math-7BBase Model=Qwen2.5-Math-7B, Training Strategy=Base, # Rollouts=–2025.110.208
GRESOBase Model=Qwen2.5-Math-1.5B, Training Strategy=GRESO, # Rollouts=2.5M2025.110.2
Random SelectionBase Model=Qwen2.5-Math-1.5B, Training Strategy=Random Selection, # Rollouts=3.0M2025.110.1667
PREPOBase Model=Qwen2.5-Math-1.5B, Training Strategy=PREPO, # Rollouts=1.1M2025.110.1667
Random SelectionBase Model=Qwen2.5-7B, Training Strategy=Random Selection, # Rollouts=716K2025.110.1641
PREPOBase Model=Qwen2.5-7B, Training Strategy=PREPO, # Rollouts=304K2025.110.1609
GRESOBase Model=Qwen2.5-7B, Training Strategy=GRESO, # Rollouts=680K2025.110.1083
Qwen2.5-Math-1.5BBase Model=Qwen2.5-Math-1.5B, Training Strategy=Base, # Rollouts=–2025.110.1021
Qwen2.5-7BBase Model=Qwen2.5-7B, Training Strategy=Base, # Rollouts=–2025.110.0417