Please refuse to answer me! Mitigating Over-Refusal in Large Language Models via Adaptive Contrastive Decoding
Year: 2026
Type: conference-paper
Abstract: Safety-aligned large language models (LLMs) often generate refusal responses to harmless queries due to the over-refusal problem.However, existing methods for mitigating overrefusal cannot maintain a low refusal ratio for harmless queries while keeping a high refusal ratio for malicious ones.In this paper, we analyze how system prompts with varying safety levels affect LLM refusal behaviors when facing over-refusal queries.A key observation is that, when LLMs suffer from the over-refusal issue, non-refusal tokens remain present in the next-token candidate list, but the model systematically fails to select them, despite the generation of refusal tokens.Based on this observation, we propose a trainingfree and model-agnostic approach, Adaptive Contrastive Decoding (AdaCD), to mitigate over-refusal while maintaining LLM safety.First, AdaCD compares the output distributions of the LLM with or without an extreme safety system prompt to refine the refusal token distribution.Second, we introduce an adaptive contrastive decoding strategy that dynamically incorporates or removes the refusal token distribution, adaptively boosting the probability of selecting refusal or non-refusal tokens.Experimental results on five benchmark datasets show that, on average, AdaCD reduces the refusal ratio for over-refusal queries by 10.35%, yet still increases the refusal ratio for malicious queries by 0.13%.
Authors: Yupeng Qi, Ziyu Lyu, Lixin Cui, Lu Bai, Feng Xia
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7166870944