SAKURA: On the Multi-hop Reasoning of Large Audio-Language Models Based on Speech and Audio Information Paper • 2505.13237 • Published 6 days ago • 2