List your company
ResearchFormerly WizWand

Language Understanding on MMLU (First-Token Accuracy)

79.7MMLU First-Token AccuracyPhi-4-14B
32.06844.43456.869.166May 21, 2025
Updated 7mo ago

Evaluation results

MethodLinks
Phi-4-14BStrategy=Prefilling2025.0579.7
Phi-4-14BStrategy=Prompt Engineering2025.0578.8
Phi-4-14BStrategy=FTP2025.0576.4
Gemma-2-9BStrategy=Prefilling2025.0572.1
Gemma-2-9BStrategy=Prompt Engineering2025.0570.2
Qwen-2-7BStrategy=Prefilling2025.0569.1
Qwen-2-7BStrategy=FTP2025.0568.9
Llama-3.1-8BStrategy=Prefilling2025.0568.4
Qwen-2-7BStrategy=Prompt Engineering2025.0568.3
Mistral-Nemo-12BStrategy=Prefilling2025.0566
Mistral-Nemo-12BStrategy=FTP2025.0565.1
Ministral-8BStrategy=Prefilling2025.0563.9
Llama-3.1-8BStrategy=Prompt Engineering2025.0563.7
Llama-3.1-8BStrategy=FTP2025.0563.1
Ministral-8BStrategy=FTP2025.0562.1
Zephyr-7BStrategy=Prefilling2025.0558.8
Mistral-Nemo-12BStrategy=Prompt Engineering2025.0558.1
Zephyr-7BStrategy=FTP2025.0557.5
Ministral-8BStrategy=Prompt Engineering2025.0557
Zephyr-7BStrategy=Prompt Engineering2025.0555.9
Gemma-7BStrategy=Prefilling2025.0552.4
Gemma-7BStrategy=Prompt Engineering2025.0548.1
Gemma-7BStrategy=FTP2025.0545.9
Gemma-2-9BStrategy=FTP2025.0533.9