List your company

Factuality and Reasoning on MMLU

55.8MMLU AccuracySpIEL
45.448.150.853.5Jun 3, 2024

Evaluation results

MethodLinks
SpIELBackbone=Llama-2 13B, SFT=true2024.0655.8
LoRABackbone=Llama-2 13B, SFT=true2024.0655.8
FullFTBackbone=Llama-2 13B, SFT=true2024.0655.6
SATBackbone=Llama-2 13B, SFT=true2024.0655.4
OriginalBackbone=Llama-2 13B, SFT=false2024.0655.3
(IA)³Backbone=Llama-2 13B, SFT=true2024.0654.3
SpIELBackbone=Llama-2 7B, SFT=true2024.0650.7
FullFTBackbone=Llama-2 7B, SFT=true2024.0650.5
SATBackbone=Llama-2 7B, SFT=true2024.0650.5
LoRABackbone=Llama-2 7B, SFT=true2024.0649.3
(IA)³Backbone=Llama-2 7B, SFT=true2024.0646.7
OriginalBackbone=Llama-2 7B, SFT=false2024.0645.8