List your company

Language Modeling on PubMed OOD from The Pile

12.61PerplexityASEntmax
12.385213.902615.4216.9374Jun 19, 2025

Evaluation results

MethodLinks
ASEntmaxContext Length=8K2025.0612.61
EntmaxContext Length=8K2025.0612.86
ASEntmaxContext Length=16K2025.0612.9
EntmaxContext Length=16K2025.0613.02
SSMaxContext Length=8K2025.0613.75
SSMaxContext Length=16K2025.0614.72
ASEntmaxContext Length=4K2025.0614.76
EntmaxContext Length=4K2025.0614.79
SSMaxContext Length=4K2025.0615.14
SoftmaxContext Length=8K2025.0615.31
SoftmaxContext Length=4K2025.0615.59
SoftmaxContext Length=16K2025.0618.23