DGoettlich/history-llms
Information hub for our project training the largest possible historical LLMs.
Information hub for our project training the largest possible historical LLMs.
Official implementation for the paper "DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Models"
Companion code for the global workspace interpretability paper
shared templates