List your company
ResearchFormerly WizWand

Math on GSM8K (Pass@1)

96.51Pass@1GT-Reward
13.455635.017856.5878.1422Nov 3, 2025Nov 30, 2025Dec 27, 2025Jan 23, 2026Feb 19, 2026Mar 18, 2026Apr 14, 2026
Updated 6mo ago

Evaluation results

MethodLinks
GT-RewardBase Model=Qwen3-8B-Base2025.1196.51
GT-RewardBase Model=Llama-3.1-8B-Instruct2025.1195.6
Self-HarmonyBase Model=Qwen3-8B-Base2025.1195.45
Co-RewardBase Model=Qwen3-8B-Base2025.1194.8
GT-RewardBase Model=Qwen3-4B-Base2025.1194.69
Self-HarmonyBase Model=Qwen3-4B-Base2025.1194.31
Majority-VotingBase Model=Qwen3-8B-Base2025.1194
GT-RewardBase Model=Llama-3.2-3B-Instruct2025.1193.7
Co-RewardBase Model=Qwen3-4B-Base2025.1193.47
Majority-VotingBase Model=Qwen3-4B-Base2025.1193.44
IntuitorBase Model=Qwen3-8B-Base2025.1192.28
Self-HarmonyBase Model=Llama-3.1-8B-Instruct2025.1191.59
RentBase Model=Qwen3-8B-Base2025.1191.2
Ling-flash base-2.0Shots=82026.0490.75
N-3-Super 120B-A12B-BaseShots=82026.0490.67
RentBase Model=Qwen3-4B-Base2025.1190.49
Self-HarmonyBase Model=Llama-3.2-3B-Instruct2025.1189.55
Co-RewardBase Model=Llama-3.1-8B-Instruct2025.1189.48
Co-RewardBase Model=Llama-3.2-3B-Instruct2025.1189.14
Majority-VotingBase Model=Llama-3.1-8B-Instruct2025.1188.78
IntuitorBase Model=Qwen3-4B-Base2025.1187.53
Self-HarmonyBase Model=Qwen3-1.7B-Base2025.1187.47
Co-RewardBase Model=Qwen3-1.7B-Base2025.1186.59
Majority-VotingBase Model=Llama-3.2-3B-Instruct2025.1185.98
GT-RewardBase Model=Qwen3-1.7B-Base2025.1185.97
Before RLBase Model=Qwen3-8B-Base2025.1184.76
Majority-VotingBase Model=Qwen3-1.7B-Base2025.1183.8
GLM-4.5 Air-BaseShots=82026.0482.6
ProSeCo SamplingCorrector Sampling=true, Number of shots=52026.0282.18
LLaDA1.5Corrector Sampling=false, Number of shots=52026.0281.12
Vanilla SFT + ReMDMCorrector Sampling=true, Training strategy=SFT, Number of shots=52026.0280.97
IntuitorBase Model=Qwen3-1.7B-Base2025.1180.25
ProSeCo SFTCorrector Sampling=false, Number of shots=52026.0279.45
LLaDA-Instruct + ReMDMCorrector Sampling=true, Number of shots=52026.0279.08
LLaDA-InstructCorrector Sampling=false, Number of shots=52026.0278.85
RentBase Model=Qwen3-1.7B-Base2025.1178.64
Vanilla SFTCorrector Sampling=false, Training strategy=SFT, Number of shots=52026.0277.48
Llama3.1-InstructCorrector Sampling=false, Number of shots=52026.0276.88
RentBase Model=Llama-3.1-8B-Instruct2025.1169.88
LLaDA-BaseCorrector Sampling=false, Number of shots=52026.0266.72
IntuitorBase Model=Llama-3.1-8B-Instruct2025.1166.25
Before RLBase Model=Qwen3-1.7B-Base2025.1165.58
Before RLBase Model=Llama-3.1-8B-Instruct2025.1160.48
Before RLBase Model=Qwen3-4B-Base2025.1155.72
IntuitorBase Model=Llama-3.2-3B-Instruct2025.1116.73
RentBase Model=Llama-3.2-3B-Instruct2025.1116.67
Before RLBase Model=Llama-3.2-3B-Instruct2025.1116.65