paper-with-me

홈 › Papers

Preemptive Detection and Steering of LLM Misalignment via Latent Reachability

2025-09-25 · Sathwik Karnik, Somil Bansal arxiv

Large language models (LLMs) are now ubiquitous in everyday tools, raising urgent safety concerns about their tendency to generate harmful content. The dominant safety approach -- reinforcement learning from human feedback (RLHF) -- effectively shapes model behavior during training but offers no safeguards at inference time, where unsafe continuations may still arise. We propose BRT-Align, a reachability-based framework that brings control-theoretic safety tools to LLM inference. BRT-Align models autoregressive generation as a dynamical system in latent space and learn a safety value function via backward reachability, estimating the worst-case evolution of a trajectory. This enables two complementary mechanisms: (1) a runtime monitor that forecasts unsafe completions several tokens in advance, and (2) a least-restrictive steering filter that minimally perturbs latent states to redirect generation away from unsafe regions. Experiments across multiple LLMs and toxicity benchmarks demonstrate that BRT-Align provides more accurate and earlier detection of unsafe continuations than baselines. Moreover, for LLM safety alignment, BRT-Align substantially reduces unsafe generations while preserving sentence diversity and coherence. Qualitative results further highlight emergent alignment properties: BRT-Align consistently produces responses that are less violent, less profane, less offensive, and less politically biased. Together, these findings demonstrate that reachability analysis provides a principled and practical foundation for inference-time LLM safety.

📄 PDF Abstract BibTeX arXiv:2509.21528

Code (0)

등록된 구현이 없습니다.

Tasks

Reinforcement Learning

Similar Papers 제목 키워드 기반

Concept Reachability in Diffusion Models: Beyond Dataset Constraints

2025-05-25 · Marta Aparicio Rodriguez, Xenia Miscouridou, Anastasia Borovykh

Despite significant advances in quality and complexity of the generations in text-to-image models, prompting does not always lead to the desired outputs. Controlling model behaviour by directly steering intermediate mode…

Preemptive Detection and Correction of Misaligned Actions in LLM Agents

2024-07-16 · Haishuo Fang, Xiaodan Zhu, Iryna Gurevych

Deploying LLM-based agents in real-life applications often faces a critical challenge: the misalignment between agents' behavior and user intent. Such misalignment may lead agents to unintentionally execute critical acti…

Action DetectionDecision Making

Uncertainty-aware Latent Safety Filters for Avoiding Out-of-Distribution Failures

2025-05-01 · Junwon Seo, Kensuke Nakamura, Andrea Bajcsy

Recent advances in generative world models have enabled classical safe control methods, such as Hamilton-Jacobi (HJ) reachability, to generalize to complex robotic systems operating directly from high-dimensional sensor …

Conformal Prediction

Activation Steering Induces Emergent Misalignment: A More Comprehensive Evaluation

2026-06-07 · Qi Cao, Jian Lou, Meiting Liu, Wenjie Feng 외 arxiv

Activation steering has emerged as a popular inference-time technique for modulating the behavior of large language models (LLMs). By constructing a steering vector from examples of a target behavior and injecting it int…

Vision-Language Introspection: Mitigating Overconfident Hallucinations in MLLMs via Interpretable Bi-Causal Steering

2026-01-08 · Shuliang Liu, Songbo Yang, Dong Fang, Sihang Jia 외 arxiv

Object hallucination critically undermines the reliability of Multimodal Large Language Models, often stemming from a fundamental failure in cognitive introspection, where models blindly trust linguistic priors over spec…