Cover Image

Over-Refusal and Representation Subspaces: A Mechanistic Analysis of Task-Conditioned Refusal in Aligned LLMs

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Maskey, Utsav, Dras, Mark, Naseem, Usman
Format:	Preprint
Published:	2026
Subjects:	Computation and Language
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

SafeConstellations: Mitigating Over-Refusals in LLMs Through Task-Aware Representation Steering
by: Maskey, Utsav, et al.
Published: (2025)

Should LLM Safety Be More Than Refusing Harmful Instructions?
by: Maskey, Utsav, et al.
Published: (2025)

Steering Over-refusals Towards Safety in Retrieval Augmented Generation
by: Maskey, Utsav, et al.
Published: (2025)

DUAL-Bench: Measuring Over-Refusal and Robustness in Vision-Language Models
by: Ren, Kaixuan, et al.
Published: (2025)

Benchmarking Large Language Models for Cryptanalysis and Side-Channel Vulnerabilities
by: Maskey, Utsav, et al.
Published: (2025)

We Think, Therefore We Align LLMs to Helpful, Harmless and Honest Before They Go Wrong
by: Kashyap, Gautam Siddharth, et al.
Published: (2025)

AlignCultura: Towards Culturally Aligned Large Language Models?
by: Kashyap, Gautam Siddharth, et al.
Published: (2026)

Fairness Evaluation and Inference Level Mitigation in LLMs
by: Nadeem, Afrozah, et al.
Published: (2025)

Steering Towards Fairness: Mitigating Political Bias in LLMs
by: Nadeem, Afrozah, et al.
Published: (2025)

Framing Political Bias in Multilingual LLMs Across Pakistani Languages
by: Nadeem, Afrozah, et al.
Published: (2025)

SHIELD: Classifier-Guided Prompting for Robust and Safer LVLMs
by: Ren, Juan, et al.
Published: (2025)

Seeing the Threat: Vulnerabilities in Vision-Language Models to Adversarial Attack
by: Ren, Juan, et al.
Published: (2025)

Beyond Over-Refusal: Scenario-Based Diagnostics and Post-Hoc Mitigation for Exaggerated Refusals in LLMs
by: Yuan, Shuzhou, et al.
Published: (2025)

When the Model Said 'No Comment', We Knew Helpfulness Was Dead, Honesty Was Alive, and Safety Was Terrified
by: Kashyap, Gautam Siddharth, et al.
Published: (2026)

Too Helpful, Too Harmless, Too Honest or Just Right?
by: Kashyap, Gautam Siddharth, et al.
Published: (2025)

CogMem: A Cognitive Memory Architecture for Sustained Multi-Turn Reasoning in Large Language Models
by: Zhang, Yiran, et al.
Published: (2025)

Can LLMs Refuse Questions They Do Not Know? Measuring Knowledge-Aware Refusal in Factual Tasks
by: Pan, Wenbo, et al.
Published: (2025)

Beyond the Black Box: Demystifying Multi-Turn LLM Reasoning with VISTA
by: Zhang, Yiran, et al.
Published: (2025)

VITAL: A New Dataset for Benchmarking Pluralistic Alignment in Healthcare
by: Shetty, Anudeex, et al.
Published: (2025)

MaiBERT: A Pre-training Corpus and Language Model for Low-Resourced Maithili Language
by: Yadav, Sumit, et al.
Published: (2025)

Mechanistic Interpretability for Large Language Model Alignment: Progress, Challenges, and Future Directions
by: Naseem, Usman
Published: (2026)

Refusal in LLMs is an Affine Function
by: Marshall, Thomas, et al.
Published: (2024)

LLMs Encode Harmfulness and Refusal Separately
by: Zhao, Jiachen, et al.
Published: (2025)

What Drives Representation Steering? A Mechanistic Case Study on Steering Refusal
by: Cheng, Stephen, et al.
Published: (2026)

Refusal Direction is Universal Across Safety-Aligned Languages
by: Wang, Xinpeng, et al.
Published: (2025)

Think Before Refusal : Triggering Safety Reflection in LLMs to Mitigate False Refusal Behavior
by: Si, Shengyun, et al.
Published: (2025)

Cannot or Should Not? Automatic Analysis of Refusal Composition in IFT/RLHF Datasets and Refusal Behavior of Black-Box LLMs
by: von Recum, Alexander, et al.
Published: (2024)

Refuse Whenever You Feel Unsafe: Improving Safety in LLMs via Decoupled Refusal Training
by: Yuan, Youliang, et al.
Published: (2024)

Linearly Decoding Refused Knowledge in Aligned Language Models
by: Shrivastava, Aryan, et al.
Published: (2025)

OR-Bench: An Over-Refusal Benchmark for Large Language Models
by: Cui, Justin, et al.
Published: (2024)

Programming Refusal with Conditional Activation Steering
by: Lee, Bruce W., et al.
Published: (2024)

How Post-Training Reshapes LLMs: A Mechanistic View on Knowledge, Truthfulness, Refusal, and Confidence
by: Du, Hongzhe, et al.
Published: (2025)

RAID: Refusal-Aware and Integrated Decoding for Jailbreaking LLMs
by: Nguyen, Tuan T., et al.
Published: (2025)

Latent Adversarial Training Improves the Representation of Refusal
by: Abbas, Alexandra, et al.
Published: (2025)

Mitigating Over-Refusal in Aligned Large Language Models via Inference-Time Activation Energy
by: Jiang, Eric Hanchen, et al.
Published: (2025)

Towards Understanding and Improving Refusal in Compressed Models via Mechanistic Interpretability
by: Chhabra, Vishnu Kabir, et al.
Published: (2025)

From Rogue to Safe AI: The Role of Explicit Refusals in Aligning LLMs with International Humanitarian Law
by: Mavi, John, et al.
Published: (2025)

Refusal Tokens: A Simple Way to Calibrate Refusals in Large Language Models
by: Jain, Neel, et al.
Published: (2024)

Beyond No: Quantifying AI Over-Refusal and Emotional Attachment Boundaries
by: Noever, David, et al.
Published: (2025)

Silenced Biases: The Dark Side LLMs Learned to Refuse
by: Himelstein, Rom, et al.
Published: (2025)