Over-Refusal

Follow this topic Ask about this topic

Over-refusal occurs when a model declines a safe, useful request because it mistakes sensitive wording, emotional language, or context for harmful intent. Researchers measure it with benchmarks that pair benign and harmful prompts, including tests in healthcare, law, cybersecurity, and tool-using agents, and assess it alongside harmful compliance. Mitigation approaches include curated fine-tuning and preference or process rewards, as well as inference-time methods such as adaptive decoding, activation steering, and semantic routing. The central challenge is to make refusals track intent, evidence, and permission across tasks and languages while preserving both helpful responses and defenses against genuinely harmful requests.

Latest papers

27 papers5 papers in the last four weeks, up 67% on the four weeks before. 0.0% of all new papers.

Revisiting Over-refusal in Safety-aligned Large Language Models: A Representation and Optimization Perspective
Oct 29, 2026Chao Chen, Pengyang Shao, Wei Qin +3Over-Refusal Refusal Behavior in Language Models
Hefei University of Technology Hefei, China · National University of Singapore Singapore, Singapore
Evaluation Awareness Beyond English: Do LLMs Act Differently When Tested in Indic Languages?
Oct 15, 2026Trizal Garg, Sibayan Mitra, Arya Karanjkar +6Evaluation Awareness in Language Models LLM Safety Evaluation
Birla Institute of Technology and Science, Pilani
Manacá-1B-Instruct-2: Reproducible Safety Alignment for a Native Brazilian-Portuguese LLM. Negative DPO Results, a Controlled Composition Scan, and a Silent Failure Mode of the LLM Judge
Oct 14, 2026Bruno Leonardo Santos Menezes, Carlos Leonardo Souza Cardoso, Fabio André Machado PortoPortuguese Language Models LLM Safety Alignment
National Laboratory for Scientific Computing (LNCC), Institute of Artificial Intelligence, Petrópolis, Brazil
EmoRSS: Mitigating Emotion-Induced Over-Refusal in Large Language Models
Oct 4, 2026Shuyi Miao, Yaojin Ma, Chenhang Cui +5LLM Alignment LLM Refusal Behavior
Beihang University · University of Science and Technology Beijing · National University of Singapore +2
OpenMTB-Audit: Exposing Over-Refusal and Clinical Expert Perspectives in LLM-Based Molecular Tumor Board Safety Evaluation
Oct 1, 2026Negin Ashrafi, Jia Luo, Stacey M. Frumm +1LLM Safety Benchmarks LLM Auditing
Department of Biomedical Data Science, Stanford University, Stanford, CA, USA · Lowe Center for Thoracic Oncology, Dana-Farber Cancer Institute, and Harvard Medical School, Boston, MA, USA
DeShortcut-Align: Decoupling Spurious Shortcuts for Robust Safety Alignment in Large Reasoning Models
Sep 28, 2026Qirui Liu, Yichen Sun, Yan Wang +4Shortcut Learning in Language Models LLM Safety Alignment
The State Key Laboratory of Blockchain and Data Security, Zhejiang University · Ant Group · Chongqing Ant Consumer Finance Co., Ltd
AgentBoundary: Counterfactual Evaluation of Safety in Tool-Using LLM Agents
Sep 27, 2026Tianzhuo Yang, Zirui Mi, Yantao Huang +4LLM Agent Evaluation LLM Agents
Peking University · Beijing Academy of Artificial Intelligence
AUDITPLAN: Commit, Then Answer for Auditable Safety Alignment
Sep 16, 2026Sai Sri Pushpa Jampani, Kshitij Mishra, Asif EkbalLLM Alignment LLM Auditing
Indian Institute of Technology Patna, Bihar, India
Mitigating LLM Over-Refusal via Dynamic Semantic Routing Calibration
Sep 6, 2026Zixuan Wang, Bingjie Zhang, He Zhao +1Mechanistic Interpretability LLM Refusal Behavior
School of Artificial Intelligence, Jilin University · CSIRO, Australia · Center of Excellence for Generative AI, King Abdullah University of Science and Technology
aiXamine: Unified Black-Box Evaluation of Cross-Dimensional Trade-offs in LLM Safety, Security, and Privacy
Aug 20, 2026Fatih Deniz, Yazan Boshmaf, Dorde Popovic +1LLM Evaluation LLM Safety Evaluation
Qatar Computing Research Institute (QCRI), HBKU, Doha, Qatar
HiRoute: Hierarchical Routed Prompt Tuning for Safety Alignment of Large Language Models
Aug 13, 2026Fangzhou Chen, Shiji Zhao, Mengyang Wang +4LLM Safety Alignment Prompt Tuning
Institute of Artificial Intelligence, Beihang University
OPIUM: Mitigating Steering Externalities and Over-Refusal via Dual Objective Latent Optimization
Jul 22, 2026Kavin Aravindan, Arihant Rastogi, Krishak Aneja +4LLM Safety Alignment Over-Refusal
IIIT Hyderabad, India.
Addressing Over-Refusal in LLMs with Competing Rewards
Jun 30, 2026Taeyoun Kim, Aviral KumarOver-Refusal LLM Safety
Carnegie Mellon University · SynthLabs
LLMs Prompted for Legal Context Object More: Overrefusal from Small On-Premises LLMs in Criminal Legal Context
Jun 23, 2026Anastasiia Kucherenko, François Brouchoud, Dimitri Percia David +1LLM Prompting Small Language Models
IEM, HEG, HES-SO Valais-Wallis Sierre, Switzerland
Measuring & Mitigating Over-Alignment for LLMs in Multilingual Criminal Law Courts
Jun 22, 2026Arthur Wuhrmann, Gaetan Stein, Daniel Brunner +1Multilingual Language Model Evaluation Legal NLP
Swiss Federal Supreme Court · HES-SO Valais-Wallis
FreoStream:Enhancing Stream Guardrails via Future-Aware Reasoning and Safety-Aligned Optimization
Jun 11, 2026Jianwei Wang, Guoyang Shen, Yanhong Wu +5LLM Guardrails Over-Refusal
South China University of Technology · BUAA
DDOR: Delta Debugging for Explainable Overrefusal Testing and Repair
Jun 2, 2026Qinyan Zhou, Peixin Zhang, Jun Sun +2LLM Guardrails LLM Refusal Behavior
Southeast University, China · School of Computing and Information Systems, Singapore Management University, Singapore · Zhejiang University, China +1
Beyond "I cannot fulfill this request": Alleviating Rigid Rejection in LLMs via Label Enhancement
May 8, 2026Ying Zhang, Congyu Qiao, Xin Geng +1LLM Safety Alignment LLM Refusal Behavior
The Refusal--Compliance Tradeoff: A Large-Scale Safety Behavior Audit of Large Language Models
May 6, 2026Alif Al Hasan, Sumon BiswasLLM Safety Benchmarks Language Model Safety Evaluation
Department of Computer and Data Sciences Case Western Reserve University Cleveland, OH, USA
Self-Mined Hardness for Safety Fine-Tuning
May 4, 2026Prakhar Gupta, Garv Shah, Donghua ZhangRefusal Behavior in Language Models Hard Negative Mining
University of Michigan
Please refuse to answer me! Mitigating Over-Refusal in Large Language Models via Adaptive Contrastive Decoding
Apr 18, 2026Yupeng Qi, Ziyu Lyu, Lixin Cui +2Language Model Safety Evaluation LLM Refusal Behavior
School of Cyber Science and Technology, Shenzhen Campus of Sun Yat-sen University · School of Information, Central University of Finance and Economics · School of Artificial Intelligence, Beijing Normal University +1
On Safety Risks in Experience-Driven Self-Evolving Agents
Apr 18, 2026Weixiang Zhao, Yichen Zhang, Yingshuo Wang +8LLM Safety Evaluation Over-Refusal
Harbin Institute of Technology · Singapore Management University · SERES
Over-Refusal and Representation Subspaces: A Mechanistic Analysis of Task-Conditioned Refusal in Aligned LLMs
Mar 29, 2026Utsav Maskey, Mark Dras, Usman NaseemLLM Interpretability LLM Refusal Behavior
Macquarie University
MOSAIC: Multi-Objective Slice-Aware Iterative Curation for Alignment
Mar 19, 2026Yipu Dou, Wang YangMulti-Objective Language Model Alignment LLM Safety Alignment
School of Cyber Science and Engineering, Southeast University, Nanjing 211189, China
Deactivating Refusal Triggers: Understanding and Mitigating Overrefusal in Safety Alignment
Mar 12, 2026Zhiyu Xue, Zimo Qi, Guangliang Liu +2LLM Safety Alignment Over-Refusal
University of California, Santa Barbara · Johns Hopkins University · Michigan State University +1
Defensive Refusal Bias: How Safety Alignment Fails Cyber Defenders
Mar 1, 2026David Campbell, Neil Kale, Udari Madhushani Sehwag +5LLM Safety Alignment LLM Refusal Behavior
Security and Policy Research Lab Scale AI · Security Engineering Scale AI
Health-ORSC-Bench: A Benchmark for Measuring Over-Refusal and Safety Completion in Health Context
Jan 25, 2026Zhihao Zhang, Liting Huang, Guanghao Wu +3LLM Safety Benchmarks Language Model Safety Evaluation
Macquarie University1 · UTS2 · MBZUAI3 +1

Oct 29, 2026J/K move · Enter open · S save

Revisiting Over-refusal in Safety-aligned Large Language Models: A Representation and Optimization Perspective

Chao Chen, Pengyang Shao, Wei Qin, Shijie Hao+2

Hefei University of Technology Hefei, China · National University of Singapore Singapore, Singapore

Over-Refusal Refusal Behavior in Language Models Safety-Preserving Fine-Tuning of Language Models

Open paper AskPDF arXiv

Over-refusal is a critical side effect of safety alignment in large language models (LLMs), causing harmless but superficially risky queries to be mistakenly rejected. Despite the effectiveness of existing mitigation methods, why safety-aligned LLMs may still reject harmless queries under correct safety supervision remains insufficiently understood. To address this gap, we analyze over-refusal through the lens of representation and optimization. We make two observations: 1) the representations of harmful and over-refusal samples are visually distinguishable at certain intermediate layers but increasingly overlap in deeper layers; and 2) harmful-sample gradients exhibit high intra-class similarity and concentrate in a low-rank subspace. Moreover, they are more strongly aligned with refusal-target gradients than with correct helpful-target gradients on over-refusal prompts. This imbalanced gradient coupling can bias over-refusal samples toward refusal, while the reduced apparent representational separation in deeper layers may make it more difficult for the model to maintain a clear boundary between harmful and over-refusal samples. To this end, we propose GOAL (Gradient Orthogonalization with contrAstive Learning), a safety-alignment framework for mitigating over-refusal. Specifically, GOAL consists of two components: 1) to prevent imbalanced gradient coupling from biasing the optimization of over-refusal samples toward refusal, we orthogonalize over-refusal gradients against the low-rank refusal subspace extracted from harmful gradients; and 2) to explicitly enhance the separation between harmful and over-refusal representations, we introduce a contrastive learning objective. Extensive experiments across multiple models and benchmarks demonstrate that GOAL effectively mitigates over-refusal while preserving safety performance and general capabilities.