Large language models (LLMs) can teach other algorithms unwanted traits, which can persist even when training data has been ...
I tried training a classifier, then found a better solution.