Refusal in Language Models Is Mediated by a Single Direction

• URL: http://arxiv.org/abs/2406.11717v3
• Date: Wed, 30 Oct 2024 18:57:07 GMT
• Title: Refusal in Language Models Is Mediated by a Single Direction
• Authors: Andy Arditi, Oscar Obeso, Aaquib Syed, Daniel Paleka, Nina Panickssery, Wes Gurnee, Neel Nanda,
• Abstract summary: We show that refusal is mediated by a one-dimensional subspace, across 13 popular open-source chat models up to 72B parameters in size. We propose a novel white-box jailbreak method that surgically disables refusal with minimal effect on other capabilities.
• Score: 4.532520427311685
• License:
• Abstract: Conversational large language models are fine-tuned for both instruction-following and safety, resulting in models that obey benign requests but refuse harmful ones. While this refusal behavior is widespread across chat models, its underlying mechanisms remain poorly understood. In this work, we show that refusal is mediated by a one-dimensional subspace, across 13 popular open-source chat models up to 72B parameters in size. Specifically, for each model, we find a single direction such that erasing this direction from the model's residual stream activations prevents it from refusing harmful instructions, while adding this direction elicits refusal on even harmless instructions. Leveraging this insight, we propose a novel white-box jailbreak method that surgically disables refusal with minimal effect on other capabilities. Finally, we mechanistically analyze how adversarial suffixes suppress propagation of the refusal-mediating direction. Our findings underscore the brittleness of current safety fine-tuning methods. More broadly, our work showcases how an understanding of model internals can be leveraged to develop practical methods for controlling model behavior.

Related papers

• Turning Logic Against Itself : Probing Model Defenses Through Contrastive Questions [51.51850981481236]
  We introduce POATE, a novel jailbreak technique that harnesses contrastive reasoning to provoke unethical responses. PoATE crafts semantically opposing intents and integrates them with adversarial templates, steering models toward harmful outputs with remarkable subtlety. To counter this, we propose Intent-Aware CoT and Reverse Thinking CoT, which decompose queries to detect malicious intent and reason in reverse to evaluate and reject harmful responses.
  arXiv  Detail & Related papers  (2025-01-03T15:40:03Z)
• Safety Alignment Backfires: Preventing the Re-emergence of Suppressed Concepts in Fine-tuned Text-to-Image Diffusion Models [57.16056181201623]
  Fine-tuning text-to-image diffusion models can inadvertently undo safety measures, causing models to relearn harmful concepts. We present a novel but immediate solution called Modular LoRA, which involves training Safety Low-Rank Adaptation modules separately from Fine-Tuning LoRA components. This method effectively prevents the re-learning of harmful content without compromising the model's performance on new tasks.
  arXiv  Detail & Related papers  (2024-11-30T04:37:38Z)
• Surgical, Cheap, and Flexible: Mitigating False Refusal in Language Models via Single Vector Ablation [29.605302471407537]
  Training a language model to be both helpful and harmless requires careful calibration of refusal behaviours. We propose a simple and surgical method for mitigating false refusal in language models via single vector ablation. Our approach is training-free and model-agnostic, making it useful for mitigating the problem of false refusal in current and future language models.
  arXiv  Detail & Related papers  (2024-10-04T13:25:32Z)
• Steering Without Side Effects: Improving Post-Deployment Control of Language Models [61.99293520621248]
  Language models (LMs) have been shown to behave unexpectedly post-deployment. We present KL-then-steer (KTS), a technique that decreases the side effects of steering while retaining its benefits. Our best method prevents 44% of jailbreak attacks compared to the original Llama-2-chat-7B model.
  arXiv  Detail & Related papers  (2024-06-21T01:37:39Z)
• Who's asking? User personas and the mechanics of latent misalignment [12.92431783194089]
  misaligned capabilities remain latent in safety-tuned models. We show that even when model generations are safe, harmful content can persist in hidden representations. We investigate why certain personas break model safeguards and find that they enable the model to form more charitable interpretations.
  arXiv  Detail & Related papers  (2024-06-17T21:15:12Z)
• Navigating the OverKill in Large Language Models [84.62340510027042]
  We investigate the factors for overkill by exploring how models handle and determine the safety of queries. Our findings reveal the presence of shortcuts within models, leading to an over-attention of harmful words like 'kill' and prompts emphasizing safety will exacerbate overkill. We introduce Self-Contrastive Decoding (Self-CD), a training-free and model-agnostic strategy, to alleviate this phenomenon.
  arXiv  Detail & Related papers  (2024-01-31T07:26:47Z)
• Gaining Wisdom from Setbacks: Aligning Large Language Models via Mistake Analysis [127.85293480405082]
  The rapid development of large language models (LLMs) has not only provided numerous opportunities but also presented significant challenges. Existing alignment methods usually direct LLMs toward the favorable outcomes by utilizing human-annotated, flawless instruction-response pairs. This study proposes a novel alignment technique based on mistake analysis, which deliberately exposes LLMs to erroneous content to learn the reasons for mistakes and how to avoid them.
  arXiv  Detail & Related papers  (2023-10-16T14:59:10Z)
• Beyond Trivial Counterfactual Explanations with Diverse Valuable Explanations [64.85696493596821]
  In computer vision applications, generative counterfactual methods indicate how to perturb a model's input to change its prediction. We propose a counterfactual method that learns a perturbation in a disentangled latent space that is constrained using a diversity-enforcing loss. Our model improves the success rate of producing high-quality valuable explanations when compared to previous state-of-the-art methods.
  arXiv  Detail & Related papers  (2021-03-18T12:57:34Z)

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.