Perch 2.0 transfers ‘whale’ to underwater tasks
Andrea BurnsGoogle DeepMindLauren HarrellGoogle ResearchBart van MerriënboerGoogle DeepMindVincent DumoulinGoogle DeepMindJenny HamerGoogle DeepMindTom DentonGoogle DeepMind · 2025
Read Perch 2.0 transfers ‘whale’ to underwater tasks by Andrea BurnsGoogle DeepMindLauren HarrellGoogle ResearchBart van MerriënboerGoogle DeepMindVincent DumoulinGoogle DeepMindJenny HamerGoogle DeepMindTom DentonGoogle DeepMind (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-03219.txt.
Description
A bird song model may also hear whales. Andrea Burns and co-authors test Perch 2.0, a bioacoustics model, on underwater sounds in this 2025 work. The model was trained on many thousands of species, from birds to insects.
Almost no marine mammal audio was in its training data. Still, the team tests it on ocean tasks using few-shot transfer, where a model adapts from few examples. Good transfer would save the cost of training new models for each habitat. It could speed up wildlife monitoring in the sea. Few-shot success would let one strong model serve both land and sea wildlife surveys.
Catalog record
| CREATOR | Andrea BurnsGoogle DeepMindLauren HarrellGoogle ResearchBart van MerriënboerGoogle DeepMindVincent DumoulinGoogle DeepMindJenny HamerGoogle DeepMindTom DentonGoogle DeepMind |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 18 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-03219