List your company
ResearchFormerly WizWand

Reasoning on ARC-E

98.1First-Token AccuracyZephyr-7B
36.2252.28568.3584.415May 21, 2025Jun 13, 2025Jul 6, 2025Jul 29, 2025Aug 21, 2025Sep 13, 2025Oct 7, 2025
Updated 4mo ago

Evaluation results

MethodLinks
Zephyr-7BStrategy=Prefilling2025.0598.1
Gemma-2-9BStrategy=Prefilling2025.0596.6
Zephyr-7BStrategy=FTP2025.0595.6
Llama-3.1-8BStrategy=Prefilling2025.0594.7
Qwen-2-7BStrategy=Prefilling2025.0594.7
Qwen-2-7BStrategy=FTP2025.0594.3
Ministral-8BStrategy=Prefilling2025.0593.7
Ministral-8BStrategy=FTP2025.0593.6
Mistral-Nemo-12BStrategy=Prefilling2025.0593.1
Mistral-Nemo-12BStrategy=FTP2025.0592.6
Llama-3.1-8BStrategy=FTP2025.0590.2
Phi-4-14BStrategy=FTP2025.0587.8
Phi-4-14BStrategy=Prefilling2025.0587.8
Gemma-7BStrategy=Prefilling2025.0586.7
Gemma-2-9BStrategy=Prompt Engineering2025.0581.2
Llama-3.1-8BStrategy=Prompt Engineering2025.0578.7
Mistral-Nemo-12BStrategy=Prompt Engineering2025.0576.5
BaseBackbone=LLaMA-2-7B, Compression Ratio=0%2025.1076.3
Ministral-8BStrategy=Prompt Engineering2025.0576.1
Qwen-2-7BStrategy=Prompt Engineering2025.0574.1
Phi-4-14BStrategy=Prompt Engineering2025.0573.1
PGSVDBackbone=LLaMA-2-7B, Compression Ratio=20%2025.1070.75
Zephyr-7BStrategy=Prompt Engineering2025.0565.7
LLM-PrunerBackbone=LLaMA-2-7B, Compression Ratio=20%2025.1064.31
Gemma-7BStrategy=Prompt Engineering2025.0558.7
Gemma-7BStrategy=FTP2025.0555.9
SliceGPTBackbone=LLaMA-2-7B, Compression Ratio=20%2025.1046.09
Gemma-2-9BStrategy=FTP2025.0538.6