AI models trained to recognize different animal sounds can now be mathematically added together to create a master classifier without sharing any raw data.
Protecting privacy and moving massive audio files between institutions is a major barrier to global wildlife monitoring. Researchers have developed a way to merge separate AI models by treating their learned knowledge as simple vectors. You can take a model that knows bird calls and add it to one that knows whale songs to create a single tool that understands both. This process does not require access to the original recordings, which keeps sensitive location data safe. This breakthrough allows scientists around the world to build a collaborative, global ear for the planet's biodiversity.
Ecologically-Constrained Task Arithmetic for Multi-Taxa Bioacoustic Classifiers Without Shared Data
arXiv · 2605.03914
Training data for bioacoustics is scattered across taxa, regions, and institutions. Centralizing it all is often infeasible. We show that independently fine-tuned BEATs encoders can be composed into a unified 661-species classifier via task vector arithmetic without sharing data. We find that bioacoustic task vectors are near-orthogonal (cosine 0.01-0.09). Their separation aligns closely with spectral distribution distance, a gradient consistent with the acoustic niche hypothesis. This geometry