REINS: Refusal-Enhanced Inhibitory Steering with Sparse Autoencoder Features

• URL: http://arxiv.org/abs/2608.28233v1
• Date: Fri, 28 Aug 2026 11:47:29 GMT
• Title: REINS: Refusal-Enhanced Inhibitory Steering with Sparse Autoencoder Features
• Abstract summary: Sparse Autoencoders (SAEs) offer a lightweight inference-time path for adapting the behavior of large language models without retraining.<n>We observe that complex wrappers can still undermine existing SAE steering methods on harmful prompts.<n>We propose Refusal-Enhanced INhibitory Steering (REINS), which suppresses harmful continuation features and enhances safe refusal features.
• Score: 42.07883859906045
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Steering with Sparse Autoencoders (SAEs) offers a lightweight inference-time path for adapting the behavior of large language models without retraining. By exposing sparse and interpretable features, SAE steering provides a promising interface for safety control that guides harmful continuations toward refusal. However, we observe that complex wrappers can still undermine existing SAE steering methods on harmful prompts. To evaluate this failure mode systematically, we construct Generalized Undercover Instruction Safety Evaluation (GUISE), a dataset of harmful prompts with complex wrappers. Existing single direction SAE steering methods do not reliably produce refusals on harmful prompts, suggesting that refusal enhancement alone can be too weak when the harmful continuation path remains active. This motivates us to propose Refusal-Enhanced INhibitory Steering (REINS), which suppresses harmful continuation features and enhances safe refusal features in the same SAE feature space. Experiments on GUISE and other datasets show that prior methods either intervene too weakly or achieve only apparent safety through collapse, while REINS substantially reduces harmful responses, markedly improves safe refusals and largely preserves general capabilities.

Related papers

• DDOR: Delta Debugging for Explainable Overrefusal Testing and Repair [17.573294728864443]
  DDOR (Delta for OverRefusal) is a fully automated and explainable framework for overrefusal testing and repair.<n>It applies delta debug to localize minimal refusal-triggering fragments (mRTFs) that provide phrase-level, explainable evidence for why a refusal occurs.<n>It produces scalable and model-specific overrefusal test suites (approximately 1K cases per model)
  arXiv  Detail & Related papers  (2026-06-02T13:07:12Z)
• Enhancing Agent Safety Judgment: Controlled Benchmark Rewriting and Analogical Reasoning for Deceptive Out-of-Distribution Scenarios [0.2578242050187029]
  ROME is a controlled benchmark-construction pipeline that rewrites known unsafe trajectories into deceptive evaluation instances.<n>ARISE is a retrieval-guided inference-time enhancement that retrieves ReAct-style analogical safety trajectories from an external analogical base and injects them as structured reasoning exemplars.
  arXiv  Detail & Related papers  (2026-05-05T00:21:00Z)
• GSAE: Graph-Regularized Sparse Autoencoders for Robust LLM Safety Steering [5.124731939041066]
  Large language models (LLMs) face critical safety challenges as they can be manipulated to generate harmful content through adversarial prompts and jailbreak attacks.<n>We introduce Graph-Regularized Sparse Autoencoders (GSAEs), which extends SAEs with a Laplacian smoothness penalty on the neuron co-activation graph.<n>We demonstrate that GSAE enables effective safety steering, assembling features into a weighted set of safety-relevant directions and controlling them with a two-stage gating mechanism.
  arXiv  Detail & Related papers  (2025-12-07T04:46:30Z)
• Oyster-I: Beyond Refusal -- Constructive Safety Alignment for Responsible Language Models [93.5740266114488]
  Constructive Safety Alignment (CSA) protects against malicious misuse while actively guiding vulnerable users toward safe and helpful results.<n>Oy1 achieves state-of-the-art safety among open models while retaining high general capabilities.<n>We release Oy1, code, and the benchmark to support responsible, user-centered AI.
  arXiv  Detail & Related papers  (2025-09-02T03:04:27Z)
• Turning the Spell Around: Lightweight Alignment Amplification via Rank-One Safety Injection [47.347413305965006]
  Safety alignment in Large Language Models (LLMs) often involves mediating internal representations to refuse harmful requests.<n>Recent research has demonstrated that these safety mechanisms can be bypassed by ablating or removing specific representational directions.<n>We propose Rank-One Safety Injection (ROSI), a white-box method that amplifies a model's safety alignment by permanently steering its activations toward the refusal-mediating subspace.
  arXiv  Detail & Related papers  (2025-08-28T13:22:33Z)
• Shape it Up! Restoring LLM Safety during Finetuning [65.75757313781104]
  Finetuning large language models (LLMs) enables user-specific customization but introduces critical safety risks.<n>We propose dynamic safety shaping (DSS), a framework that uses fine-grained safety signals to reinforce learning from safe segments of a response while suppressing unsafe content.<n>We present STAR-DSS, guided by STAR scores, that robustly mitigates finetuning risks and delivers substantial safety improvements across diverse threats, datasets, and model families.
  arXiv  Detail & Related papers  (2025-05-22T18:05:16Z)
• SCANS: Mitigating the Exaggerated Safety for LLMs via Safety-Conscious Activation Steering [56.92068213969036]
  Safety alignment is indispensable for Large Language Models (LLMs) to defend threats from malicious instructions.<n>Recent researches reveal safety-aligned LLMs prone to reject benign queries due to the exaggerated safety issue.<n>We propose a Safety-Conscious Activation Steering (SCANS) method to mitigate the exaggerated safety concerns.
  arXiv  Detail & Related papers  (2024-08-21T10:01:34Z)
• Refuse Whenever You Feel Unsafe: Improving Safety in LLMs via Decoupled Refusal Training [67.30423823744506]
  We introduce a novel approach, Decoupled Refusal Training (DeRTa), designed to empower LLMs to refuse compliance to harmful prompts at any response position.<n>DeRTa incorporates two novel components: (1) Maximum Likelihood Estimation with Harmful Response Prefix, which trains models to recognize and avoid unsafe content by appending a segment of harmful response to the beginning of a safe response, and (2) Reinforced Transition Optimization (RTO), which equips models with the ability to transition from potential harm to safety refusal consistently throughout the harmful response sequence.
  arXiv  Detail & Related papers  (2024-07-12T09:36:33Z)
• On Prompt-Driven Safeguarding for Large Language Models [172.13943777203377]
  We find that in the representation space, the input queries are typically moved by safety prompts in a "higher-refusal" direction. Inspired by these findings, we propose a method for safety prompt optimization, namely DRO. Treating a safety prompt as continuous, trainable embeddings, DRO learns to move the queries' representations along or opposite the refusal direction, depending on their harmfulness.
  arXiv  Detail & Related papers  (2024-01-31T17:28:24Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.