一作ACL 2026Findings· Findings of ACL
Characterizing Selective Refusal Bias in Large Language Models
Adel Khorramrouz · Sharon Levy
摘要
Safety guardrails in large language models (LLMs) are developed to prevent malicious users from generating toxic content at a large scale. However, these measures can inadvertently introduce or reflect new biases, as LLMs may refuse to generate harmful content targeting some demographic groups an...