List your company
ResearchFormerly WizWand

Question Answering on MMLU

88.7AccuracyQwen 3 VL 32B Instruct
20.340838.087955.83573.5821May 11, 2023Nov 12, 2023May 16, 2024Nov 18, 2024May 22, 2025Nov 24, 2025May 29, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Qwen 3 VL 32B InstructParameters=32B2025.1288.7
GPT-4Number of shots=5-shot2023.1186.5
Qwen 3 32BThinking=No, Parameters=32B2025.1285.8
Qwen 2.5 32BParameters=32B2025.1284.6
Qwen 3 VL 8B Inststage=Instruct2025.1283.6
Olmo 3.1 32B InstructStage=DPO2025.1281.9
Olmo 3.1 32B InstructStage=Final Instruct 3.12025.1280.9
Qwen 3 8Bstage=Instruct2025.1280.4
Olmo 3.1 32B InstructStage=SFT2025.1279
Qwen 2.5 7Bstage=Instruct2025.1277.2
OLMo 2 32BParameters=32B2025.1277.1
Gemma 2 27BParameters=27B2025.1276.1
Gemma 3 27BParameters=27B2025.1274.6
FalconModel Size=180B2023.1170.6
Falcon-180BNumber of shots=5-shot2023.1170.6
Mixtral 8x7B2024.0170.6
Apertus 70BParameters=70B2025.1270.2
GPT-3.5Number of shots=5-shot2023.1170
GPT-3.52024.0170
LLaMA 2 70B2024.0169.9
PaLM2023.1169.3
Olmo 3 7B Instructstage=DPO2025.1269.1
Olmo 3 7B Instructstage=Final Instruct2025.1269.1
ExpGraphBackbone Model=Llama-3.1-8B-Instruct (Large LLM), Method Category=LLM-Centric Experience Learning Baselines2026.0569
LLaMA-2Model Size=70B2023.1168.9
RCOBase Model=LLaMA-3-8B-Instruct2025.0667.74
LLaMA-3-70B-Instruct2025.0667.39
Olmo 3 7B Instructstage=SFT2025.1267.1
LLaDA-8B-InstructSettings=Official scoreboard evaluation settings2026.0165.4
CE OnlyTraining=CE Only (Matched Baseline)2026.0165.1
RCOBase Model=UltraCM-13B2025.0664.42
CTC+CETraining=CTC-trained2026.0164.1
Granite 3.3 8B Inststage=Instruct2025.1263.5
RCOBase Model=LLaMA-2-7B-Chat2025.0663.17
Apertus 8B Inststage=Instruct2025.1262.7
LLaMA-2Model Size=34B2023.1162.6
RCOBase Model=LLaMA-2-13B-Chat2025.0662.48
OLMo 2 7B Inststage=Instruct2025.1261.6
DPCOBase Model=LLaMA-2-7B-Chat2025.0660.46
RCOBase Model=Auto-J-13B2025.0660.37
DPCOBase Model=UltraCM-13B2025.0660.03
LLAMA-2-70B-Chat2025.0660.02
ExpGraphBackbone Model=Llama-3.2-3B-Instruct (Small LLM), Method Category=LLM-Centric Experience Learning Baselines2026.0560
DPCOBase Model=LLaMA-2-13B-Chat2025.0659.42
LLaMA-3-8B-Instruct2025.0659.41
DPCOBase Model=LLaMA-3-8B-Instruct2025.0659.21
DPCOBase Model=Auto-J-13B2025.0658.47
LLaMA-2-13B-Chat2025.0657.33
UltraCM-13B2025.0657.19
FalconModel Size=40B2023.1157
LLaMA-2-7B-Chat2025.0656.96
STABLEPROMPTPrompt LM=Gemma-7B, Target LM=Gemma-7B2026.0255.8
GFLOWPOPrompt LM=Gemma-7B, Target LM=Gemma-7B2026.0255.6
LLaMA-2Model Size=13B2023.1154.8
Self-refinement2025.0654.79
ProTEGIPrompt LM=Gemma-7B, Target LM=Gemma-7B2026.0253.5
Auto-J-13B2025.0652.9
Aligner2025.0652.56
Initial Answer2025.0652.48
APEPrompt LM=Gemma-7B, Target LM=Gemma-7B2026.0252.1
LLaMA-2Model Size=7B2023.1145.3
OLTQAFull Config=True2023.0536.14
UnifiedQA2023.0528.77
FalconModel Size=7B2023.1128
DCDM (MoE)Scale=1.5B, Active Params=1.2B (2.8B), In-context examples=0-shot2026.0526.1
DCDMScale=1.5B, Active Params=1.5B, In-context examples=0-shot2026.0525.59
BDLMScale=1.5B, Active Params=1.5B, In-context examples=0-shot2026.0525.22
DCDM (MoE)Scale=0.5B, Active Params=0.4B (0.8B), In-context examples=0-shot2026.0525.09
MDLMScale=1.5B, Active Params=1.5B, In-context examples=0-shot2026.0524.49
DCDMScale=0.5B, Active Params=0.5B, In-context examples=0-shot2026.0524.32
BDLMScale=0.5B, Active Params=0.5B, In-context examples=0-shot2026.0523.53
MDLMScale=0.5B, Active Params=0.5B, In-context examples=0-shot2026.0523.24
Softmaxzero-shot=true2025.0122.97
LSSARzero-shot=true, p=152025.0122.97