Reinforcement Learning Fine-tuning of Language Models is Biased Towards More Extractable Features

Add code
Nov 07, 2023

Share this with someone who'll enjoy it:

View paper onarxiv icon

Share this with someone who'll enjoy it: