Research

Paper

AI LLM February 23, 2026

BarrierSteer: LLM Safety via Learning Barrier Steering

Authors

Thanh Q. Tran, Arun Verma, Kiwan Wong, Bryan Kian Hsiang Low, Daniela Rus, Wei Xiao

Abstract

Despite the state-of-the-art performance of large language models (LLMs) across diverse tasks, their susceptibility to adversarial attacks and unsafe content generation remains a major obstacle to deployment, particularly in high-stakes settings. Addressing this challenge requires safety mechanisms that are both practically effective and supported by rigorous theory. We introduce BarrierSteer, a novel framework that formalizes response safety by embedding learned non-linear safety constraints directly into the model's latent representation space. BarrierSteer employs a steering mechanism based on Control Barrier Functions (CBFs) to efficiently detect and prevent unsafe response trajectories during inference with high precision. By enforcing multiple safety constraints through efficient constraint merging, without modifying the underlying LLM parameters, BarrierSteer preserves the model's original capabilities and performance. We provide theoretical results establishing that applying CBFs in latent space offers a principled and computationally efficient approach to enforcing safety. Our experiments across multiple models and datasets show that BarrierSteer substantially reduces adversarial success rates, decreases unsafe generations, and outperforms existing methods.

Metadata

arXiv ID: 2602.20102
Provider: ARXIV
Primary Category: cs.LG
Published: 2026-02-23
Fetched: 2026-02-24 04:38

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2602.20102v1</id>\n    <title>BarrierSteer: LLM Safety via Learning Barrier Steering</title>\n    <updated>2026-02-23T18:19:46Z</updated>\n    <link href='https://arxiv.org/abs/2602.20102v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2602.20102v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Despite the state-of-the-art performance of large language models (LLMs) across diverse tasks, their susceptibility to adversarial attacks and unsafe content generation remains a major obstacle to deployment, particularly in high-stakes settings. Addressing this challenge requires safety mechanisms that are both practically effective and supported by rigorous theory. We introduce BarrierSteer, a novel framework that formalizes response safety by embedding learned non-linear safety constraints directly into the model's latent representation space. BarrierSteer employs a steering mechanism based on Control Barrier Functions (CBFs) to efficiently detect and prevent unsafe response trajectories during inference with high precision. By enforcing multiple safety constraints through efficient constraint merging, without modifying the underlying LLM parameters, BarrierSteer preserves the model's original capabilities and performance. We provide theoretical results establishing that applying CBFs in latent space offers a principled and computationally efficient approach to enforcing safety. Our experiments across multiple models and datasets show that BarrierSteer substantially reduces adversarial success rates, decreases unsafe generations, and outperforms existing methods.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.LG'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.AI'/>\n    <published>2026-02-23T18:19:46Z</published>\n    <arxiv:comment>This paper introduces SafeBarrier, a framework that enforces safety in large language models by steering their latent representations with control barrier functions during inference, reducing adversarial and unsafe outputs</arxiv:comment>\n    <arxiv:primary_category term='cs.LG'/>\n    <author>\n      <name>Thanh Q. Tran</name>\n    </author>\n    <author>\n      <name>Arun Verma</name>\n    </author>\n    <author>\n      <name>Kiwan Wong</name>\n    </author>\n    <author>\n      <name>Bryan Kian Hsiang Low</name>\n    </author>\n    <author>\n      <name>Daniela Rus</name>\n    </author>\n    <author>\n      <name>Wei Xiao</name>\n    </author>\n  </entry>"
}