AINeutralarXiv – CS AI · Mar 176/10
🧠
Evaluation of Audio Language Models for Fairness, Safety, and Security
Researchers introduce a structural taxonomy and unified evaluation framework for Audio Large Language Models (ALLMs) to assess fairness, safety, and security. The study reveals systematic differences in how ALLMs handle audio versus text inputs, with FSS behavior closely tied to acoustic information integration methods.