mlsanityData to 7 Oct 2026

Surgical, Cheap, and Flexible: Mitigating False Refusal in Language Models via Single Vector Ablation

arXiv 2410.03415 ↗First posted 4 Oct 2024cs.CLcited by 48

Not on the map. Why?

Cited by 22 in LLM safety and security, but placed by its own abstract, which didn’t fit.

Abstract (arXiv)

Training a language model to be both helpful and harmless requires careful calibration of refusal behaviours: Models should refuse to follow malicious instructions or give harmful advice (e.g."how do I kill someone?"), but they should not refuse safe requests, even if they superficially resemble unsafe ones (e.g. "how do I kill a Python process?"). Avoiding such false refusal, as prior work has shown, is challenging even for highly-capable language models. In this paper, we propose a simple and surgical method for mitigating false refusal in language models via single vector ablation. For a given model, we extract a false refusal vector and show that ablating this vector reduces false refusal rate while preserving the model's safety and general capabilities. We also show that our approach can be used for fine-grained calibration of model safety. Our approach is training-free and model-agnostic, making it useful for mitigating the problem of false refusal in current and future language models.

Key facts. Cited by 48 papers in the map (data through 7 Oct 2026). Builds on 38 of its 49 references that are in the map. First posted on arXiv 4 Oct 2024.

Builds on

38 of 49 references.

References · 38

  1. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
  2. 2020Measuring Massive Multitask Language Understandingcited by 7,768
  3. 2018Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challengecited by 4,710
  4. 2022Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedbackcited by 3,634
  5. 2016Pointer Sentinel Mixture Modelscited by 2,992
  6. 2021TruthfulQA: Measuring How Models Mimic Human Falsehoodscited by 2,919
  7. 2023Universal and Transferable Adversarial Attacks on Aligned Language Modelscited by 2,914
  8. 2024HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusalcited by 1,330
  9. 2023Inference-Time Intervention: Eliciting Truthful Answers from a Language Modelcited by 1,259
  10. 2023Representation Engineering: A Top-Down Approach to AI Transparencycited by 1,239
  11. 2023Steering Llama 2 via Contrastive Activation Additioncited by 1,009
  12. 2024Refusal in Language Models Is Mediated by a Single Directioncited by 960
26 more
  1. 2023AutoDAN: Generating Stealthy Jailbreak Prompts on Aligned Large Language Modelscited by 738
  2. 2023Safe RLHF: Safe Reinforcement Learning from Human Feedbackcited by 630
  3. 2023DecodingTrust: A Comprehensive Assessment of Trustworthiness in GPT Modelscited by 505
  4. 2024JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Modelscited by 494
  5. 2023XSTest: A Test Suite for Identifying Exaggerated Safety Behaviours in Large Language Modelscited by 477
  6. 2024WildGuard: Open One-Stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMscited by 460
  7. 2023Catastrophic Jailbreak of Open-source LLMs via Exploiting Generationcited by 439
  8. 2023SmoothLLM: Defending Large Language Models Against Jailbreaking Attackscited by 434
  9. 2023Safety-Tuned LLaMAs: Lessons From Improving the Safety of Large Language Models that Follow Instructionscited by 377
  10. 2024SafeDecoding: Defending against Jailbreak Attacks via Safety-Aware Decodingcited by 233
  11. 2024SORRY-Bench: Systematically Evaluating Large Language Model Safety Refusalcited by 224
  12. 2024OR-Bench: An Over-Refusal Benchmark for Large Language Modelscited by 222
  13. 2020HateCheck: Functional Tests for Hate Speech Detection Modelscited by 215
  14. 2023An Overview of Catastrophic AI Riskscited by 214
  15. 2023LoRA Fine-tuning Efficiently Undoes Safety Training in Llama 2-Chat 70Bcited by 183
  16. 2020Recipes for Safety in Open-domain Chatbotscited by 168
  17. 2024Evaluating Frontier Models for Dangerous Capabilitiescited by 121
  18. 2024On Prompt-Driven Safeguarding for Large Language Modelscited by 109
  19. 2023Cognitive Overload: Jailbreaking Large Language Models with Overloaded Logical Thinkingcited by 78
  20. 2021On the Safety of Conversational Models: Taxonomy, Dataset, and Benchmarkcited by 73
  21. 2023SimpleSafetyTests: a Test Suite for Identifying Critical Safety Risks in Large Language Modelscited by 72
  22. 2024Navigating the OverKill in Large Language Modelscited by 40
  23. 2024Automatic Pseudo-Harmful Prompt Generation for Evaluating False Refusals in Large Language Modelscited by 35
  24. 2024From Theft to Bomb-Making: The Ripple Effect of Unlearning in Defending Against Jailbreak Attackscited by 29
  25. 2023LoFT: Local Proxy Fine-tuning For Improving Transferability Of Adversarial Attacks Against Large Language Modelcited by 21
  26. 2024From Representational Harms to Quality-of-Service Harms: A Case Study on Llama 2 Safety Safeguardscited by 17

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 1; 2025: 20; 2026: 27.

48 papers.

Sort:

By area:

  1. 2024Steering Language Model Refusal with Sparse Autoencoderscited by 89
  2. 2025FalseReject: A Resource for Improving Contextual Safety and Mitigating Over-Refusals in LLMs via Structured Reasoningcited by 53
  3. 2025AlphaSteer: Learning Refusal Steering with Principled Null-Space Constraintcited by 45
  4. 2025AdaSteer: Your Aligned LLM is Inherently an Adaptive Jailbreak Defendercited by 37
  5. 2025Refusal Direction is Universal Across Safety-Aligned Languagescited by 37
  6. 2025Just Enough Shifts: Mitigating Over-Refusal in Aligned Language Models with Targeted Representation Fine-Tuningcited by 22
  7. 2026Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Modelscited by 22
  8. 2025X-Boundary: Establishing Exact Safety Boundary to Shield LLMs from Multi-Turn Jailbreaks without Compromising Usabilitycited by 20
  9. 2025Steering the CensorShip: Uncovering Representation Vectors for LLM "Thought" Controlcited by 14
  10. 2025MMPB: It's Time for Multi-Modal Personalizationcited by 14
  11. 2025Think Before Refusal : Triggering Safety Reflection in LLMs to Mitigate False Refusal Behaviorcited by 12
  12. 2025No for Some, Yes for Others: Persona Prompts and Other Sources of False Refusal in Language Modelscited by 8
36 more
  1. 2025Do We Know What LLMs Don't Know? A Study of Consistency in Knowledge Probingcited by 7
  2. 2026What Drives Representation Steering? A Mechanistic Case Study on Steering Refusalcited by 7
  3. 2025LoRA is All You Need for Safety Alignment of Reasoning LLMscited by 6
  4. 2025Turning the Spell Around: Lightweight Alignment Amplification via Rank-One Safety Injectioncited by 5
  5. 2025Unveiling the "Fairness Seesaw": Discovering and Mitigating Gender and Race Bias in Vision-Language Modelscited by 4
  6. 2025Mitigating Over-Refusal in Aligned Large Language Models via Inference-Time Activation Energycited by 4
  7. 2026Minimizing Collateral Damage in Activation Steeringcited by 4
  8. 2025MCP Safety Training: Learning to Refuse Falsely Benign MCP Exploits using Improved Preference Alignmentcited by 3
  9. 2025Toward Preference-aligned Large Language Models via Residual-based Model Steeringcited by 3
  10. 2026Steering to Say No: Configurable Refusal via Activation Steering in Vision Language Modelscited by 3
  11. 2026Discern Truth from Falsehood: Reducing Over-Refusal via Contrastive Refinementcited by 3
  12. 2025The Safety Gap Toolkit: Evaluating Hidden Dangers of Open-Source Modelscited by 2
  13. 2025Dynamic Evaluation for Oversensitivity in LLMscited by 2
  14. 2025Task Vector Bases: A Unified and Scalable Framework for Compressed Task Arithmeticcited by 1
  15. 2026Over-Refusal and Representation Subspaces: A Mechanistic Analysis of Task-Conditioned Refusal in Aligned LLMscited by 1
  16. 2026FineSteer: A Unified Framework for Fine-Grained Inference-Time Steering in Large Language Modelscited by 1
  17. 2026Safety Geometry Collapse in Multimodal LLMs and Adaptive Drift Correctioncited by 1
  18. 2026Palette: A Modular, Controllable, and Efficient Framework for On-demand Authorized Safety Alignment Relaxation in LLMscited by 1
  19. 2026Beyond Refusal: A Same-Lineage Study of Aligned and Abliterated LLMs for Vulnerability Analysiscited by 1
  20. 2026BabelSteering: Multilingual Safety Alignment via English Steering Vectorscited by 1
  21. 2026Does Fine-Tuning Undo Activation Steering? Behavioural Recovery Without Weight-Edit Reversalcited by 1
  22. 2026C-DeltaTheta: Circuit-Restricted Weight Arithmetic for Selective Refusalcited by 0
  23. 2026Residual Stream Analysis of Overfitting And Structural Disruptionscited by 0
  24. 2026A Systematic Study of Training-Free Methods for Trustworthy Large Language Modelscited by 0
  25. 2026Please refuse to answer me! Mitigating Over-Refusal in Large Language Models via Adaptive Contrastive Decodingcited by 0
  26. 2026Steering Beyond the Support: Adversarial Training on Unsupervised Jailbroken Activation Simulationcited by 0
  27. 2026DDOR: Delta Debugging for Explainable Overrefusal Testing and Repaircited by 0
  28. 2026LoMC: Localized Multidirectional Correction for Refusal Suppression in Routed Foundation Modelscited by 0
  29. 2026AOR-Bench: Do Large Audio Language Models Over-Refuse Pseudo-Harmful Queries?cited by 0
  30. 2026Pretraining Curricula Enable Selective Fine-tuningcited by 0
  31. 2026Robust Critics: Defending LLMs Against Multi-Turn Attackscited by 0
  32. 2026Making Your LLMs More Objective: Stabilizing LLM Safety Behavior Across Traits with Trait-Invariant Safety Tuningcited by 0
  33. 2026Abliteration Mitigation via Refusal Aliasescited by 0
  34. 2026Refuse without Refusal: A Structural Analysis of Safety-Tuning Responses for Reducing False Refusals in Language Modelscited by 0
  35. 2026Decoy Direction Optimization: A Post-Hoc Defense Against LLM Abliterationcited by 0
  36. 2026Mitigating LLM Over-Refusal via Dynamic Semantic Routing Calibrationcited by 0