MetaPerch: Learning from metadata for bioacoustics foundation models

Mustafa Chasmai, Vincent Dumoulin, Jenny Hamer
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:13066-13088, 2026.

Abstract

Bioacoustic foundation models rely on large-scale citizen science platforms like Xeno-Canto for geographically and ecologically diverse data. Recent work has shown that supervision alone can produce SotA species detection models when trained on this large-scale data—however, there remains unutilized potential in the form of recording metadata readily available within these community-driven data hubs. In this work, we explore the use of metadata—such as location and time—as auxiliary supervision signals, allowing the model to leverage species-metadata correlations in its learned representation. Auxiliary metadata losses provide additional information beyond vocalizations alone that can encourage a richer, more robust representation that generalizes better to species distribution and acoustic domain shifts—important challenges for deployment in real-world passive acoustic monitoring (PAM) settings. We introduce MetaPerch, a new foundation model that achieves strong species identification performance across multiple challenging domains and present an extensive empirical study of the effects of 9 diverse metadata sources on 17 bioacoustic datasets.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-chasmai26a, title = {{M}eta{P}erch: Learning from metadata for bioacoustics foundation models}, author = {Chasmai, Mustafa and Dumoulin, Vincent and Hamer, Jenny}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {13066--13088}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/chasmai26a/chasmai26a.pdf}, url = {https://proceedings.mlr.press/v306/chasmai26a.html}, abstract = {Bioacoustic foundation models rely on large-scale citizen science platforms like Xeno-Canto for geographically and ecologically diverse data. Recent work has shown that supervision alone can produce SotA species detection models when trained on this large-scale data—however, there remains unutilized potential in the form of recording metadata readily available within these community-driven data hubs. In this work, we explore the use of metadata—such as location and time—as auxiliary supervision signals, allowing the model to leverage species-metadata correlations in its learned representation. Auxiliary metadata losses provide additional information beyond vocalizations alone that can encourage a richer, more robust representation that generalizes better to species distribution and acoustic domain shifts—important challenges for deployment in real-world passive acoustic monitoring (PAM) settings. We introduce MetaPerch, a new foundation model that achieves strong species identification performance across multiple challenging domains and present an extensive empirical study of the effects of 9 diverse metadata sources on 17 bioacoustic datasets.} }
Endnote
%0 Conference Paper %T MetaPerch: Learning from metadata for bioacoustics foundation models %A Mustafa Chasmai %A Vincent Dumoulin %A Jenny Hamer %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-chasmai26a %I PMLR %P 13066--13088 %U https://proceedings.mlr.press/v306/chasmai26a.html %V 306 %X Bioacoustic foundation models rely on large-scale citizen science platforms like Xeno-Canto for geographically and ecologically diverse data. Recent work has shown that supervision alone can produce SotA species detection models when trained on this large-scale data—however, there remains unutilized potential in the form of recording metadata readily available within these community-driven data hubs. In this work, we explore the use of metadata—such as location and time—as auxiliary supervision signals, allowing the model to leverage species-metadata correlations in its learned representation. Auxiliary metadata losses provide additional information beyond vocalizations alone that can encourage a richer, more robust representation that generalizes better to species distribution and acoustic domain shifts—important challenges for deployment in real-world passive acoustic monitoring (PAM) settings. We introduce MetaPerch, a new foundation model that achieves strong species identification performance across multiple challenging domains and present an extensive empirical study of the effects of 9 diverse metadata sources on 17 bioacoustic datasets.
APA
Chasmai, M., Dumoulin, V. & Hamer, J.. (2026). MetaPerch: Learning from metadata for bioacoustics foundation models. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:13066-13088 Available from https://proceedings.mlr.press/v306/chasmai26a.html.

Related Material