arXiv:2604.17299 · v4

Cat-DPO: Category-Adaptive Safety Alignment

首次提交 2026-04-19 15:29

版本更新 2026-11-08 04:04

雷达观测 2026-11-10 11:14

摘要

Aligning large language models with human preferences must balance two competing goals: responding helpfully to legitimate requests and reliably refusing harmful ones. Most preference-based safety alignment methods collapse safety into a single scalar that is applied uniformly to every preference pair. The result is a model that appears safe on average but remains relatively unsafe in a minority of harm categories. We cast safety alignment as a per-category constrained optimization problem and derive Cat-DPO, a direct-preference-optimization algorithm with a separate adaptive safety margin for each harm category. The margin tightens when the model still produces unsafe responses in a category and relaxes once the model catches up, so the training signal tracks each category's current difficulty rather than averaging across a single global rate. Across two LLM backbones and six preference-learning baselines, Cat-DPO improves aggregate helpfulness and harmlessness and compresses per-category safety variance and the best-to-worst gap, offering a drop-in per-category refinement of direct preference safety alignment.

作者备注：Neurips 2026