1 paper
Matthew Kowal, Goncalo Paulo, Louis Jaburi +6
As large language models are increasingly trained and fine-tuned, practitioners need methods to identify which training data drive specific behaviors, particularly unintended ones.…