https://www.lesswrong.com/posts/qmQFHCgCyEEjuy5a7/lora-fine-tuning-efficiently-undoes-safety-training-from (NN sparsity, preference learning, AI safety)
https://www.lesswrong.com/posts/qmQFHCgCyEEjuy5a7/lora-fine-tuning-efficiently-undoes-safety-training-from
View External Link: