1 paper
Alexander Matt Turner, Lisa Thiergart, Gavin Leech +4
Prompt engineering and finetuning aim to maximize language model performance on a given metric (like toxicity reduction). However, these methods do not fully elicit a model's capab…