ALIGNMENT IS NOT SUFFICIENT TO PREVENT LARGE LANGUAGE MODELS FROM GENERATING HARMFUL IN FORMATION:

Last updated