@inproceedings{sseu-bench,
title={Can Large Audio Language Models Understand Audio Well? Speech, Scene and Events Understanding Benchmark for LALMs},
author={Yin, Han and Choi, Jung-Woo},
booktitle={ICASSP 2026-2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
pages={1--5},
year={2026},
organization={IEEE}
}
Recently, Large Audio Language Models (LALMs) have progressed rapidly, demonstrating their strong efficacy in universal audio understanding through cross-modal integration. To evaluate LALMs' audio understanding performance, researchers have proposed different benchmarks. However, key aspects for real-world interactions are underexplored in existing benchmarks, i.e., audio signals typically contain both speech and non-speech components, and energy levels of these components can vary significantly across different scenarios. Moreover, most benchmarks do not consider the joint understanding of speech, scene, and events within the same audio clip. In this work, we introduce SSEU-Bench, the first versatile audio understanding benchmark that explicitly accounts for energy differences between speech and non-speech audio, with both independent and joint understanding settings for speech, scene, and events.
Recently, Large Audio Language Models (LALMs) have progressed rapidly, demonstrating their strong efficacy in universal audio understanding through cross-modal integration. To evaluate LALMs' audio understanding performance, researchers have proposed different benchmarks. However, key aspects for real-world interactions are underexplored in existing benchmarks, i.e., audio signals typically contain both speech and non-speech components, and energy levels of these components can vary significantly across different scenarios. Moreover, most benchmarks do not consider the joint understanding of speech, scene, and events within the same audio clip. In this work, we introduce SSEU-Bench, the first versatile audio understanding benchmark that explicitly accounts for energy differences between speech and non-speech audio, with both independent and joint understanding settings for speech, scene, and events.
Recently, Large Audio Language Models (LALMs) have progressed rapidly, demonstrating their strong efficacy in universal audio understanding through cross-modal integration. To evaluate LALMs' audio understanding performance, researchers have proposed different benchmarks. However, key aspects for real-world interactions are underexplored in existing benchmarks, i.e., audio signals typically contain both speech and non-speech components, and energy levels of these components can vary significantly across different scenarios. Moreover, most benchmarks do not consider the joint understanding of speech, scene, and events within the same audio clip. In this work, we introduce SSEU-Bench, the first versatile audio understanding benchmark that explicitly accounts for energy differences between speech and non-speech audio, with both independent and joint understanding settings for speech, scene, and events.
