RAS: Measuring LLM Safety Through Refusal Alignment

Abstract

Safety evaluation of large language models (LLMs) is commonly performed by querying models with unsafe or jailbreak prompts and judging whether their outputs violate a safety policy. Although useful, output-level evaluation is expensive, sensitive to judge choice, and easily tied to fixed question banks. We propose **SafeVec**, a white-box evaluation procedure that measures safety from internal representations rather than generated answers. **SafeVec** first extracts layer-wise refusal direction

Categories

Framework mappings

Suggested from the entry's categories.

@misc{huang2026ras,
  title = {{RAS: Measuring LLM Safety Through Refusal Alignment}},
  author = {Chang-Chieh Huang and Yan-Lun Chen and Chia-Mu Yu and Wei-Bin Lee},
  year = {2026},
  month = jun,
  eprint = {2606.25750},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2606.25750}
}