Preserving Safety During Sequential Fine-Tuning: A Factorial and Geometric Analysis of Replay
Benign fine-tuning can erode safety alignment in large language models, and sequential adaptation substantially compounds this risk. Prior work mitigates this erosion by replaying harmful-prompt refusals and benign-compliance examples, while other work instead matches base-model logits via…
By George Ruellan, Diana Benavides-PradoarXiv cs.LG
Read the full story
Originally published by arXiv cs.LG on Oct 30, 2026.
Comments
Sign in to join the discussion.