Restoring General and Improving Domain-Specific Safety in Fine-Tuned LLMs Without Retraining
SPECTRA is proposed, a training-free, weight-level method that restores refusal behavior by transferring the base model's refusal direction into the fine-tuned model's weights via an activation-aware, low-rank update, requiring no gradient-based optimization or domain-specific safety data.
V. Chhabra, Ding Zhu, Zhi-Hui Zhu et al.
· 0 citations