RepIt: Steering Language Models with Concept-Specific Refusal Vectors

• URL: http://arxiv.org/abs/2509.13281v4
• Date: Mon, 20 Oct 2025 21:21:12 GMT
• Title: RepIt: Steering Language Models with Concept-Specific Refusal Vectors
• Abstract summary: RepIt is a framework for isolating concept-specific representations.<n>It selectively suppresses refusal on targeted concepts while preserving refusal elsewhere.<n>It produces models that answer WMD-related questions while still scoring as safe on standard benchmarks.
• Score: 44.49401545656479
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: While activation steering in large language models (LLMs) is a growing area of research, methods can often incur broader effects than desired. This motivates isolation of purer concept vectors to enable targeted interventions and understand LLM behavior at a more granular level. We present RepIt, a simple and data-efficient framework for isolating concept-specific representations. Across five frontier LLMs, RepIt enables precise interventions: it selectively suppresses refusal on targeted concepts while preserving refusal elsewhere, producing models that answer WMD-related questions while still scoring as safe on standard benchmarks. We further show that the corrective signal localizes to just 100-200 neurons and that robust target representations can be extracted from as few as a dozen examples on a single A6000. This efficiency raises a dual concern: manipulations can be performed with modest compute and data to extend to underrepresented data-scarce topics while evading existing benchmarks. By disentangling refusal vectors with RepIt, this work demonstrates that targeted interventions can counteract overgeneralization, laying the foundation for more granular control of model behavior.

Related papers

• Can LLMs Reliably Self-Report Adversarial Prefills, and How? [9.80193616788089]
  We show that large language models (LLMs) exhibit introspective capability on benign tasks.<n>We examine how reliably a model can recognize that its own prior response was elicited by an adversarial prefill attack.
  arXiv  Detail & Related papers  (2026-06-22T17:56:30Z)
• FRISM: Fine-Grained Reasoning Injection via Subspace-Level Model Merging for Vision-Language Models [20.47311573790516]
  We propose FRISM (Fine-grained Reasoning Injection via Subspace-level model Merging), a fine-grained reasoning injection framework based on subspace-level model merging.<n>Experiments demonstrate that FRISM effectively improves reasoning capabilities without compromising the model's original visual capabilities.
  arXiv  Detail & Related papers  (2026-01-29T02:36:19Z)
• NeuroFilter: Privacy Guardrails for Conversational LLM Agents [50.75206727081996]
  This work addresses the computational challenge of enforcing privacy for agentic Large Language Models (LLMs)<n>NeuroFilter is a guardrail framework that operationalizes contextual integrity by mapping norm violations to simple directions in the model's activation space.<n>A comprehensive evaluation across over 150,000 interactions, covering models from 7B to 70B parameters, illustrates the strong performance of NeuroFilter.
  arXiv  Detail & Related papers  (2026-01-21T05:16:50Z)
• Reasoning with Confidence: Efficient Verification of LLM Reasoning Steps via Uncertainty Heads [104.9566359759396]
  We propose a lightweight alternative for step-level reasoning verification based on data-driven uncertainty scores.<n>Our findings suggest that the internal states of LLMs encode their uncertainty and can serve as reliable signals for reasoning verification.
  arXiv  Detail & Related papers  (2025-11-09T03:38:29Z)
• Do Stop Me Now: Detecting Boilerplate Responses with a Single Iteration [0.0]
  Large Language Models (LLMs) often expend significant computational resources generating boilerplate responses.<n>We propose a simple yet highly effective method for detecting such responses after only a single generation step.
  arXiv  Detail & Related papers  (2025-10-26T13:43:56Z)
• Self-Consistency as a Free Lunch: Reducing Hallucinations in Vision-Language Models via Self-Reflection [71.8243083897721]
  Vision-language models often hallucinate details, generating non-existent objects or inaccurate attributes that compromise output reliability.<n>We present a novel framework that leverages the model's self-consistency between long responses and short answers to generate preference pairs for training.
  arXiv  Detail & Related papers  (2025-09-27T10:37:11Z)
• Rethinking Human-Object Interaction Evaluation for both Vision-Language Models and HOI-Specific Methods [33.074167753966314]
  We introduce a new benchmarking dataset that reformulates HOI detection as a multiple-answer multiple-choice task.<n>Our results show that large VLMs already surpass state-of-the-art HOI-specific methods across most metrics.
  arXiv  Detail & Related papers  (2025-08-26T07:30:53Z)
• Just Enough Shifts: Mitigating Over-Refusal in Aligned Language Models with Targeted Representation Fine-Tuning [19.823784666021822]
  ACTOR minimizes over-refusals by leveraging internal activation patterns from diverse queries.<n>ACTOR precisely identifies and adjusts the activation components that trigger refusals, providing stronger control over the refusal mechanism.
  arXiv  Detail & Related papers  (2025-07-06T05:47:04Z)
• DEAL: Disentangling Transformer Head Activations for LLM Steering [19.770342907146965]
  We propose a principled causal-attribution framework for identifying behavior-relevant attention heads in transformers.<n>For each head, we train a vector-quantized autoencoder (VQ-AE) on its attention activations.<n>We assess the behavioral relevance of each head by the separability of VQ-AE encodings for behavior-aligned versus behavior-violating responses.
  arXiv  Detail & Related papers  (2025-06-10T02:16:50Z)
• More is Less: The Pitfalls of Multi-Model Synthetic Preference Data in DPO Safety Alignment [80.04449725137177]
  Direct Preference Optimization (DPO) has emerged as a simple, yet effective alternative to reinforcement learning from human feedback.<n>Our study reveals a striking, safety-specific phenomenon associated with DPO alignment.<n>Using solely self-generated responses for both chosen and rejected pairs significantly outperforms configurations that incorporate responses from stronger models.
  arXiv  Detail & Related papers  (2025-04-03T00:36:40Z)
• UC-MOA: Utility-Conditioned Multi-Objective Alignment for Distributional Pareto-Optimality [52.49062565901046]
  Reinforcement Learning from Human Feedback (RLHF) has become a cornerstone for aligning large language models with human values.<n>Existing approaches struggle to capture the multi-dimensional, distributional nuances of human preferences.<n>We introduce Utility-Conditioned Multi-Objective Alignment (UC-MOA), a novel framework that overcomes these limitations.
  arXiv  Detail & Related papers  (2025-03-10T09:52:42Z)
• Reinforcing Thinking through Reasoning-Enhanced Reward Models [6.636512424910708]
  Large Language Models (LLMs) exhibit great potential in complex multi-step reasoning through inference-time thinking.<n>LLMs struggle with deciding when to stop thinking due to limited self-awareness about their knowledge boundaries.<n>This work addresses these challenges by distilling the LLM's own reasoning processes into synthetic behavioral data.
  arXiv  Detail & Related papers  (2024-12-31T04:50:15Z)
• ReEval: Automatic Hallucination Evaluation for Retrieval-Augmented Large Language Models via Transferable Adversarial Attacks [91.55895047448249]
  This paper presents ReEval, an LLM-based framework using prompt chaining to perturb the original evidence for generating new test cases. We implement ReEval using ChatGPT and evaluate the resulting variants of two popular open-domain QA datasets. Our generated data is human-readable and useful to trigger hallucination in large language models.
  arXiv  Detail & Related papers  (2023-10-19T06:37:32Z)
• Self-Supervised Neuron Segmentation with Multi-Agent Reinforcement Learning [53.00683059396803]
  Mask image model (MIM) has been widely used due to its simplicity and effectiveness in recovering original information from masked images. We propose a decision-based MIM that utilizes reinforcement learning (RL) to automatically search for optimal image masking ratio and masking strategy. Our approach has a significant advantage over alternative self-supervised methods on the task of neuron segmentation.
  arXiv  Detail & Related papers  (2023-10-06T10:40:46Z)
• ReWOO: Decoupling Reasoning from Observations for Efficient Augmented Language Models [32.95155349925248]
  We propose a modular paradigm ReWOO that detaches the reasoning process from external observations, thus significantly reducing token consumption. We show that ReWOO achieves 5x token efficiency and 4% accuracy improvement on HotpotQA, a multi-step reasoning benchmark. Our illustrative work offloads reasoning ability from 175B GPT3.5 into 7B LLaMA, demonstrating the significant potential for truly efficient and scalable ALM systems.
  arXiv  Detail & Related papers  (2023-05-23T00:16:48Z)

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.