List your company
ResearchFormerly WizWand

Multi-task Language Understanding on MMLU-Pro (pass@1)

92.85Pass@1GT-Reward
6.58228.978551.37573.7715Aug 25, 2025Sep 17, 2025Oct 10, 2025Nov 3, 2025Nov 26, 2025Dec 19, 2025Jan 12, 2026
Updated 6mo ago

Evaluation results

MethodLinks
GT-RewardBase Model=Qwen3-4B-Base2025.1192.85
GT-RewardBase Model=Qwen3-8B-Base2025.1192.85
GT-RewardBase Model=Llama-3.1-8B-Instruct2025.1191.42
GT-RewardBase Model=Qwen3-1.7B-Base2025.1185.71
Self-HarmonyBase Model=Qwen3-8B-Base2025.1177.68
Majority-VotingBase Model=Qwen3-8B-Base2025.1177.58
GT-RewardBase Model=Llama-3.2-3B-Instruct2025.1172.85
RentBase Model=Qwen3-8B-Base2025.1169.91
Self-HarmonyBase Model=Qwen3-4B-Base2025.1167.68
RentBase Model=Qwen3-4B-Base2025.1166.16
IntuitorBase Model=Qwen3-4B-Base2025.1164.55
Qwen2.5-7B-InstructTraining Pipeline=PSFT → GRPO2025.0863.65
Qwen2.5-7B-InstructTraining Pipeline=SFT → GRPO2025.0862.61
Llama3.1-8B-InstructTraining Pipeline=PSFT → GRPO2025.0860.92
IntuitorBase Model=Qwen3-8B-Base2025.1160.54
Qwen2.5-7B-InstructTraining Pipeline=PSFT2025.0859.18
Qwen2.5-7B-InstructTraining Pipeline=SFT2025.0858.98
SFT + GRPOBackbone=Qwen2.5-7B-Instruct2026.0157.8
Co-RewardBase Model=Qwen3-8B-Base2025.1157.59
GRPOBackbone=Qwen2.5-7B-Instruct2026.0157
Llama3.1-8B-InstructTraining Pipeline=PSFT2025.0856.58
SCR (Ours)Backbone=Qwen2.5-7B-Instruct2026.0156.3
Self-RefineBackbone=Qwen2.5-7B-Instruct2026.0155.9
BaseBackbone=Qwen2.5-7B-Instruct2026.0155.8
Llama3.1-8B-InstructTraining Pipeline=SFT → GRPO2025.0854.39
Self-HarmonyBase Model=Qwen3-1.7B-Base2025.1153.66
Majority-VotingBase Model=Qwen3-4B-Base2025.1152.95
Co-RewardBase Model=Qwen3-4B-Base2025.1151.79
SCR-Stage IBackbone=Qwen2.5-7B-Instruct2026.0151.5
Llama3.1-8B-InstructTraining Pipeline=SFT2025.0850.29
Before RLBase Model=Qwen3-8B-Base2025.1150.09
Self-HarmonyBase Model=Llama-3.1-8B-Instruct2025.1150
SCR-SFTBackbone=Qwen2.5-7B-Instruct2026.0149.1
Co-RewardBase Model=Qwen3-1.7B-Base2025.1147.14
GRPOBackbone=Llama3.1-8B-Instruct2026.0145.5
Majority-VotingBase Model=Llama-3.1-8B-Instruct2025.1145.36
SFT + GRPOBackbone=Qwen2.5-3B-Instruct2026.0145.3
Majority-VotingBase Model=Qwen3-1.7B-Base2025.1144.82
GRPOBackbone=Qwen2.5-3B-Instruct2026.0144.6
SCR (Ours)Backbone=Qwen2.5-3B-Instruct2026.0144.5
Self-HarmonyBase Model=Llama-3.2-3B-Instruct2025.1144.29
Before RLBase Model=Llama-3.1-8B-Instruct2025.1143.75
Self-RefineBackbone=Qwen2.5-3B-Instruct2026.0143.5
SCR (Ours)Backbone=Llama3.1-8B-Instruct2026.0143.2
SFT + GRPOBackbone=Llama3.1-8B-Instruct2026.0143
Co-RewardBase Model=Llama-3.1-8B-Instruct2025.1142.77
BaseBackbone=Llama3.1-8B-Instruct2026.0142.4
SCR-Stage IBackbone=Llama3.1-8B-Instruct2026.0142.2
SCR-Stage IBackbone=Qwen2.5-3B-Instruct2026.0141.3
RentBase Model=Llama-3.1-8B-Instruct2025.1140.8
IntuitorBase Model=Llama-3.1-8B-Instruct2025.1140
BaseBackbone=Qwen2.5-3B-Instruct2026.0139.6
SCR-SFTBackbone=Llama3.1-8B-Instruct2026.0139.1
SCR-SFTBackbone=Qwen2.5-3B-Instruct2026.0138.4
IntuitorBase Model=Llama-3.2-3B-Instruct2025.1134.64
RentBase Model=Llama-3.2-3B-Instruct2025.1134.38
Before RLBase Model=Llama-3.2-3B-Instruct2025.1134.11
Co-RewardBase Model=Llama-3.2-3B-Instruct2025.1134.02
Majority-VotingBase Model=Llama-3.2-3B-Instruct2025.1131.43
IntuitorBase Model=Qwen3-1.7B-Base2025.1131.25
Before RLBase Model=Qwen3-4B-Base2025.1127.59
RentBase Model=Qwen3-1.7B-Base2025.1118.04
Before RLBase Model=Qwen3-1.7B-Base2025.1116.61
Self-RefineBackbone=Llama3.1-8B-Instruct2026.019.9