List your company
ResearchFormerly WizWand

Commonsense Reasoning on CQA

79.8First-Token AccuracyPhi-4-14B
37.99248.84659.770.554May 21, 2025
Updated 7mo ago

Evaluation results

MethodLinks
Phi-4-14BStrategy=Prompt Engineering2025.0579.8
Phi-4-14BStrategy=Prefilling2025.0579.6
Qwen-2-7BStrategy=FTP2025.0579.4
Qwen-2-7BStrategy=Prompt Engineering2025.0579.4
Qwen-2-7BStrategy=Prefilling2025.0579.4
Gemma-2-9BStrategy=Prefilling2025.0579
Gemma-2-9BStrategy=Prompt Engineering2025.0578.7
Llama-3.1-8BStrategy=Prompt Engineering2025.0577.2
Llama-3.1-8BStrategy=Prefilling2025.0576.5
Mistral-Nemo-12BStrategy=Prefilling2025.0575.4
Ministral-8BStrategy=Prefilling2025.0575
Ministral-8BStrategy=FTP2025.0574.3
Mistral-Nemo-12BStrategy=FTP2025.0574.1
Llama-3.1-8BStrategy=FTP2025.0572.7
Phi-4-14BStrategy=FTP2025.0572.5
Ministral-8BStrategy=Prompt Engineering2025.0569.6
Zephyr-7BStrategy=Prefilling2025.0569
Mistral-Nemo-12BStrategy=Prompt Engineering2025.0568.8
Gemma-7BStrategy=Prefilling2025.0568.6
Zephyr-7BStrategy=FTP2025.0566
Gemma-7BStrategy=Prompt Engineering2025.0564.7
Zephyr-7BStrategy=Prompt Engineering2025.0554.4
Gemma-7BStrategy=FTP2025.0551.6
Gemma-2-9BStrategy=FTP2025.0539.6