Steering Safely or Off a Cliff? Rethinking Specificity and Robustness in Inference-Time Interventions

• URL: http://arxiv.org/abs/2602.06256v1
• Date: Thu, 05 Feb 2026 23:14:05 GMT
• Title: Steering Safely or Off a Cliff? Rethinking Specificity and Robustness in Inference-Time Interventions
• Authors: Navita Goyal, Hal Daumé,
• Abstract summary: We show that while steering achieves high efficacy and largely maintains general and control specificity, it consistently fails to preserve robustness specificity.<n>Our work provides the first systematic evaluation of specificity in model steering, showing that standard efficacy and specificity checks are insufficient.
• Score: 2.977664945581083
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Model steering, which involves intervening on hidden representations at inference time, has emerged as a lightweight alternative to finetuning for precisely controlling large language models. While steering efficacy has been widely studied, evaluations of whether interventions alter only the intended property remain limited, especially with respect to unintended changes in behaviors related to the target property. We call this notion specificity. We propose a framework that distinguishes three dimensions of specificity: general (preserving fluency and unrelated abilities), control (preserving related control properties), and robustness (preserving control properties under distribution shifts). We study two safety-critical use cases: steering models to reduce overrefusal and faithfulness hallucinations, and show that while steering achieves high efficacy and largely maintains general and control specificity, it consistently fails to preserve robustness specificity. In the case of overrefusal steering, for example, all steering methods reduce overrefusal without harming general abilities and refusal on harmful queries; however, they substantially increase vulnerability to jailbreaks. Our work provides the first systematic evaluation of specificity in model steering, showing that standard efficacy and specificity checks are insufficient, because without robustness evaluation, steering methods may appear reliable even when they compromise model safety.

Related papers

• AMPS: Adaptive Modality Preference Steering via Functional Entropy [66.69992693275061]
  We introduce an instance-aware diagnostic metric that quantifies each modality's information contribution and reveals sample-specific susceptibility to steering.<n> Experimental results show that our instance-aware steering outperforms conventional steering in modulating modality preference.
  arXiv  Detail & Related papers  (2026-02-13T02:29:06Z)
• Steering Externalities: Benign Activation Steering Unintentionally Increases Jailbreak Risk for Large Language Models [62.16655896700062]
  Activation steering is a technique to enhance the utility of Large Language Models (LLMs)<n>We show that it unintentionally introduces critical and under-explored safety risks.<n>Experiments reveal that these interventions act as a force multiplier, creating new vulnerabilities to jailbreaks and increasing attack success rates to over 80% on standard benchmarks.
  arXiv  Detail & Related papers  (2026-02-03T12:32:35Z)
• Self-Guard: Defending Large Reasoning Models via enhanced self-reflection [54.775612141528164]
  Self-Guard is a lightweight safety defense framework for Large Reasoning Models.<n>It bridges the awareness-compliance gap, achieving robust safety performance without compromising model utility.<n>Self-Guard exhibits strong generalization across diverse unseen risks and varying model scales.
  arXiv  Detail & Related papers  (2026-01-31T13:06:11Z)
• Character as a Latent Variable in Large Language Models: A Mechanistic Account of Emergent Misalignment and Conditional Safety Failures [70.48661957773449]
  Emergent Misalignment refers to a failure mode in which fine-tuning large language models on narrowly scoped data induces broadly misaligned behavior.<n>Across multiple domains and model families, we find that fine-tuning models on data exhibiting specific character-level dispositions induces substantially stronger and more transferable misalignment than incorrect-advice fine-tuning.
  arXiv  Detail & Related papers  (2026-01-30T15:28:42Z)
• Agentic Uncertainty Quantification [76.94013626702183]
  We propose a unified Dual-Process Agentic UQ (AUQ) framework that transforms verbalized uncertainty into active, bi-directional control signals.<n>Our architecture comprises two complementary mechanisms: System 1 (Uncertainty-Aware Memory, UAM), which implicitly propagates verbalized confidence and semantic explanations to prevent blind decision-making; and System 2 (Uncertainty-Aware Reflection, UAR), which utilizes these explanations as rational cues to trigger targeted inference-time resolution only when necessary.
  arXiv  Detail & Related papers  (2026-01-22T07:16:26Z)
• LatentGuard: Controllable Latent Steering for Robust Refusal of Attacks and Reliable Response Generation [4.29885665563186]
  LATENTGUARD is a framework that combines behavioral alignment with supervised latent space control for interpretable and precise safety steering.<n>Our results show significant improvements in both safety controllability and response interpretability without compromising utility.
  arXiv  Detail & Related papers  (2025-09-24T07:31:54Z)
• Beyond Prompt Engineering: Robust Behavior Control in LLMs via Steering Target Atoms [71.85633762642125]
  The vast number of parameters in models often results in highly intertwined internal representations.<n>Recent research has explored the use of sparse autoencoders (SAE) to disentangle knowledge in high-dimensional spaces for steering.<n>We propose Steering Target Atoms (STA), a novel method that isolates and manipulates disentangled knowledge components to enhance safety.
  arXiv  Detail & Related papers  (2025-05-23T17:59:18Z)
• Predicting Safety Misbehaviours in Autonomous Driving Systems using Uncertainty Quantification [8.213390074932132]
  This paper evaluates different uncertainty quantification methods from the deep learning domain for the anticipatory testing of safety-critical misbehaviours.<n>We compute uncertainty scores as the vehicle executes, following the intuition that high uncertainty scores are indicative of unsupported runtime conditions.<n>In our study, we conducted an evaluation of the effectiveness and computational overhead associated with two uncertainty quantification methods, namely MC- Dropout and Deep Ensembles, for misbehaviour avoidance.
  arXiv  Detail & Related papers  (2024-04-29T10:28:28Z)

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.