PaperScope
LIVE · 2026-09-09 05:40 UTC

Suan: Rectifying Direct Preference Safety Alignment in Large Language Models

Oleksandr Cherednichenko, Roman Klypa

Latestcs.CLcs.LGcs.AIcs.CV
arXiv ID
2609.08634 v1
Category
Submitted
2026-09-08

Abstract

Integrating robust safety guardrails into Large Language Models (LLMs) is essential for delivering helpful yet harmless responses. While proprietary systems exhibit reliable safety controls, their underlying methodologies and trade-offs remain largely undisclosed. Achieving comparable security in open-weight models remains a persistent challenge, as post-trained variants frequently suffer from over-refusal and degraded general quality. To overcome these drawbacks, we introduce Suan, a novel preference optimization algorithm. Unlike existing methods, we formulate the optimization objective directly at the gradient level, bypassing the standard variational derivation. As a result, we obtain more interpretable and robust training dynamics. Extensive evaluations across a diverse suite of competitive baselines and benchmarks demonstrate that Suan achieves superior safety alignment while fully preserving response utility.

arXiv abs page · PDF · same-day batch