List your company
ResearchFormerly WizWand

Mathematical Reasoning on Gaokao

80AccuracyPass@8 (Upper Bound)
44.84853.97463.172.226Jan 19, 2026Jan 21, 2026Jan 23, 2026Jan 25, 2026Jan 27, 2026Jan 29, 2026Feb 1, 2026
Updated 8mo ago

Evaluation results

MethodLinks
Pass@8 (Upper Bound)Policy Model=Qwen2.5-14B-Instruct, Strategy=Best-of-8, Bound=Upper Bound2026.0180
Pass@8 (Upper Bound)Policy Model=Qwen2.5-7B-Instruct2026.0177.4
Skywork-PRM-Qwen2.5-7BTraining Samples=N/A, Aggregation Method=Mean, Policy Model=Qwen2.5-14B-Instruct, Strategy=Best-of-82026.0174.5
Qwen2.5-Math-PRM-7BTraining Samples=1500K, Aggregation Method=Mean, Policy Model=Qwen2.5-14B-Instruct, Strategy=Best-of-82026.0173.5
Majority Vote@8Policy Model=Qwen2.5-14B-Instruct, Strategy=Majority Voting, Samples=82026.0173.2
Qwen2.5-Math-7B-NAITTraining Samples=128K, Aggregation Method=Mean, Policy Model=Qwen2.5-14B-Instruct, Strategy=Best-of-82026.0173.2
Qwen2.5-Math-7B-MCRDTraining Samples=128K, Aggregation Method=Mean, Policy Model=Qwen2.5-14B-Instruct, Strategy=Best-of-82026.0172.9
RLHFlow-PRM-DeepSeek-8BTraining Samples=253K, Aggregation Method=Mean, Policy Model=Qwen2.5-14B-Instruct, Strategy=Best-of-82026.0172.7
EurusPRM-Stage2Training Samples=693K, Aggregation Method=Sum, Policy Model=Qwen2.5-14B-Instruct, Strategy=Best-of-82026.0172.2
Skywork-PRM-Qwen2.5-7BPolicy Model=Qwen2.5-7B-Instruct, Training Samples=-, Aggregation Method=Mean2026.0172.2
DPO (Random)Base Model=Qwen2.5-7B-Instruct2026.0271.4
SAGEBase Model=Qwen2.5-7B-Instruct2026.0271.4
EurusPRM-Stage2Policy Model=Qwen2.5-7B-Instruct, Training Samples=693K, Aggregation Method=Sum2026.0171.4
Qwen2.5-Math-PRM-7BPolicy Model=Qwen2.5-7B-Instruct, Training Samples=1500K, Aggregation Method=Mean2026.0171.4
Qwen2.5-Math-7B-NAITPolicy Model=Qwen2.5-7B-Instruct, Training Samples=128K, Aggregation Method=Mean2026.0171.3
EurusPRM-Stage1Training Samples=463K, Aggregation Method=Min, Policy Model=Qwen2.5-14B-Instruct, Strategy=Best-of-82026.0171.2
RLHFlow-PRM-Mistral-8BTraining Samples=273K, Aggregation Method=Mean, Policy Model=Qwen2.5-14B-Instruct, Strategy=Best-of-82026.0170.9
DPO (Full)Base Model=Qwen2.5-7B-Instruct2026.0270.6
EurusPRM-Stage1Policy Model=Qwen2.5-7B-Instruct, Training Samples=463K, Aggregation Method=Min-Max2026.0170.3
Majority Vote@8Policy Model=Qwen2.5-7B-Instruct2026.0170.1
VanillaBase Model=Qwen2.5-7B-Instruct2026.0269.9
Math-Shepherd-PRM-7BTraining Samples=445K, Aggregation Method=Mean, Policy Model=Qwen2.5-14B-Instruct, Strategy=Best-of-82026.0169.6
Pass@8 (Upper Bound)Policy Model=Qwen2.5-3B-Instruct, Evaluation Protocol=Pass@82026.0169.6
GreedyPolicy Model=Qwen2.5-14B-Instruct, Strategy=Greedy Search2026.0169.3
Qwen2.5-Math-7B-MCTraining Samples=128K, Aggregation Method=Mean, Policy Model=Qwen2.5-14B-Instruct, Strategy=Best-of-82026.0169.1
Qwen2.5-Math-7B-MCRDPolicy Model=Qwen2.5-7B-Instruct, Training Samples=128K, Aggregation Method=Mean2026.0168.6
RLHFlow-PRM-DeepSeek-8BPolicy Model=Qwen2.5-7B-Instruct, Training Samples=253K, Aggregation Method=Mean2026.0168.3
GreedyPolicy Model=Qwen2.5-7B-Instruct2026.0167
Math-Shepherd-PRM-7BPolicy Model=Qwen2.5-7B-Instruct, Training Samples=445K, Aggregation Method=Mean2026.0167
RLHFlow-PRM-Mistral-8BPolicy Model=Qwen2.5-7B-Instruct, Training Samples=273K, Aggregation Method=Mean2026.0167
Qwen2.5-Math-7B-MCPolicy Model=Qwen2.5-7B-Instruct, Training Samples=128K, Aggregation Method=Mean2026.0166.8
Majority Vote@8Policy Model=Qwen2.5-3B-Instruct, Evaluation Protocol=Majority Vote@82026.0166.5
Qwen2.5-Math-PRM-7BPolicy Model=Qwen2.5-3B-Instruct, Evaluation Protocol=Best-of-8, Training Samples=1500K, Aggregation Method=Mean2026.0163.6
Skywork-PRM-Qwen2.5-7BPolicy Model=Qwen2.5-3B-Instruct, Evaluation Protocol=Best-of-8, Aggregation Method=Mean2026.0163.1
Qwen2.5-Math-7B-NAITPolicy Model=Qwen2.5-3B-Instruct, Evaluation Protocol=Best-of-8, Training Samples=128K, Aggregation Method=Mean2026.0162.8
EurusPRM-Stage2Policy Model=Qwen2.5-3B-Instruct, Evaluation Protocol=Best-of-8, Training Samples=693K, Aggregation Method=Sum2026.0160.8
EurusPRM-Stage1Policy Model=Qwen2.5-3B-Instruct, Evaluation Protocol=Best-of-8, Training Samples=463K, Aggregation Method=Min-Max2026.0160.2
GreedyPolicy Model=Qwen2.5-3B-Instruct, Evaluation Protocol=Greedy2026.0160
Qwen2.5-Math-7B-MCRDPolicy Model=Qwen2.5-3B-Instruct, Evaluation Protocol=Best-of-8, Training Samples=128K, Aggregation Method=Mean2026.0159.9
RLHFlow-PRM-Mistral-8BPolicy Model=Qwen2.5-3B-Instruct, Evaluation Protocol=Best-of-8, Training Samples=273K, Aggregation Method=Mean2026.0159.5
RLHFlow-PRM-DeepSeek-8BPolicy Model=Qwen2.5-3B-Instruct, Evaluation Protocol=Best-of-8, Training Samples=253K, Aggregation Method=Mean2026.0159.2
Math-Shepherd-PRM-7BPolicy Model=Qwen2.5-3B-Instruct, Evaluation Protocol=Best-of-8, Training Samples=445K, Aggregation Method=Mean2026.0158.7
Qwen2.5-Math-7B-MCPolicy Model=Qwen2.5-3B-Instruct, Evaluation Protocol=Best-of-8, Training Samples=128K, Aggregation Method=Mean2026.0158.4
SAGEBase Model=Qwen2.5-3B-Instruct2026.0258.23
DPO (Full)Base Model=Qwen2.5-3B-Instruct2026.0256.9
VanillaBase Model=Qwen2.5-3B-Instruct2026.0256.4
DPO (Random)Base Model=Qwen2.5-3B-Instruct2026.0256.4
SAGEBase Model=Qwen2.5-1.5B-Instruct2026.0250.4
DPO (Random)Base Model=Qwen2.5-1.5B-Instruct2026.0248.6
DPO (Full)Base Model=Qwen2.5-1.5B-Instruct2026.0247.3
VanillaBase Model=Qwen2.5-1.5B-Instruct2026.0246.2