SEABED: SouthEast Asian Benchmark for Evaluating Audio Reasoning
This work introduces SEABED, an audio-first question answering dataset designed to benchmark language and audio reasoning models on SEA speech, and evaluates not only answer accuracy, but also whether models'stated reasoning is grounded in the audio evidence.
Harshit Rajgarhia, Asif Shaik, R. Lokesh et al.
· 0 citations