$\tau$-Multilingual: Benchmarking Voice Agents Across Languages
Across 4,500 full-duplex calls and five voice configurations, Spanish, Portuguese, and Hindi remain within 3.2 task-completion points of English, but Korean and Mandarin fall by 14.7 and 8.4 points, respectively.
Soham Ray, Edgard dos Santos Paiva, Rubén Valenzuela et al.
· 0 citations