List your company
ResearchFormerly WizWand

Language Modeling on The Pile non-copyrighted (test)

0.557BPBSIFT
0.529640.714320.8991.08368Oct 10, 2024
Updated 8mo ago

Evaluation results

MethodLinks
SIFTEvaluation Protocol=Test-Time FT, Base Model=Llama-3.2, Parameters=3B2024.100.557
SIFTEvaluation Protocol=Test-Time FT, Base Model=Phi-3, Parameters=3.8B2024.100.595
SIFTEvaluation Protocol=Test-Time FT, Base Model=Llama-3.2, Parameters=1B2024.100.606
Gemma-2Parameters=27B2024.100.629
Llama-3.2Parameters=3B2024.100.64
Phi-3Parameters=14B2024.100.651
Phi-3.5-MoEParameters=16x3.8B2024.100.656
GPT-3Parameters=175B2024.100.666
Gemma-2Parameters=9B2024.100.67
Phi-3Parameters=7B2024.100.678
Phi-3Parameters=3.8B2024.100.679
Phi-3.5Parameters=3.8B2024.100.69
Llama-3.2Parameters=1B2024.100.697
Gemma-2Parameters=2B2024.100.721
Llama-3.2-InstructParameters=3B2024.100.737
SIFTEvaluation Protocol=Test-Time FT, Base Model=GPT-2, Parameters=774M2024.100.762
Llama-3.2-InstructParameters=1B2024.100.807
SIFTEvaluation Protocol=Test-Time FT, Base Model=GPT-2, Parameters=124M2024.100.862
GPT-2Parameters=774M2024.101.093
GPT-2Parameters=124M2024.101.241