Kirill

@kskblv

CS PhD student @NorthwesternCS. Previously: pre-doc @CAAI_Booth, BA @NES_Moscow

following
followers

Fine-tuning may correct both too little and too much LLM output diversity

A paper's author says LLM diversity depends on the dataset and post-training method. Models can sometimes be more diverse than the target distribution, rather than showing the under-diversity known as mode collapse.

1 Source, 43d ago, first seen 43d ago

Featured in

Fine-tuning may correct both too little and too much LLM output diversity

Featured in

Fine-tuning may correct both too little and too much LLM output diversity