AR&D: A Framework for Retrieving and Describing Concepts for Interpreting AudioLLMs

Created by Townim Faisal, Ta Duc Huy, Siqi Pan, Jeremy Stoddard, and Zhibin Liao.

This paper is accepted as ORAL in In International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2026 (top 3% conference in signal processing).

Paper Link: https://arxiv.org/abs/2602.22253

Code availability: The code associated with this work is not publicly available, as it was developed under collaboration with Dolby Laboratories and falls under their IP policies. For questions regarding access, please contact Dolby Laboratories.

Citation

@inproceedings{ard,
	title={AR\&D: A Framework for Retrieving and Describing Concepts for Interpreting AudioLLMs},
    author={Chowdhury, Townim Faisal and Huy, Ta Duc and Pan, Siqi and Stoddard, Jeremy and Liao, Zhibin},
	year = 2026,
	booktitle = {International Conference on Acoustics, Speech, and Signal Processing (ICASSP)}
}