List your company
ResearchFormerly WizWand

Language Understanding on MMLU (5-shot Accuracy)

78.33MMLU 5-shot AccuracyLP-SFT
54.212460.473766.73572.9963Jul 6, 2026
Updated 3mo ago

Evaluation results

MethodLinks
LP-SFTBackbone=Qwen3-14B2026.0778.33
ASFTBackbone=Qwen3-14B2026.0777.94
EAFTBackbone=Qwen3-14B2026.0777.76
DFTBackbone=Qwen3-14B2026.0777.4
CEBackbone=Qwen3-14B2026.0777.39
GEMBackbone=Qwen3-14B2026.0777.13
LP-SFTBackbone=Qwen3-4B2026.0769.95
CEBackbone=Qwen3-4B2026.0769.69
ASFTBackbone=Qwen3-4B2026.0769.41
GEMBackbone=Qwen3-4B2026.0769.15
DFTBackbone=Qwen3-4B2026.0769.09
EAFTBackbone=Qwen3-4B2026.0768.99
LP-SFTBackbone=Llama3.1-8B2026.0761.94
CEBackbone=Llama3.1-8B2026.0761.6
GEMBackbone=Llama3.1-8B2026.0760.55
ASFTBackbone=Llama3.1-8B2026.0760.22
EAFTBackbone=Llama3.1-8B2026.0760
DFTBackbone=Llama3.1-8B2026.0755.14