Backend
Do bioacoustic foundation models hear through the noise ? SNR and generalization in underwater detection
arXiv cs.SD0 views
arXiv:2610.33068v1 Announce Type: new Abstract: Bioacoustic foundation models are claimed to generalize across taxa and recording conditions, but are trained mainly on terrestrial archives of focal, high-quality recordings and evaluated on benchmarks emphasizing architecture over data conditions. We evaluate four such models as frozen encoders on an underwater detection task, using linear probing on continuously recorded data from the Kerguelen Plateau. Four datasets of identical size were built to differ only in the signal-to-noise ratio (SNR) of their positive segments, and probes were trained and evaluated across all combinations. Evaluation SNR accounts for more variation in average precision (AP) than model choice, and performance peaks when training and evaluation share a noise regime --- the configuration least representative of a continuous deployment. Training on mixed SNR redistributes performance toward low-SNR conditions without improving it overall, and AP remains low throughout. In light of these results, we call for evaluation standards grounded in realistic noise and deployment conditions rather than curated, high-quality recordings.
Source：https://arxiv.org/abs/2610.33068PreviousDistributional statistics of nonequilibriumNextMeasuring Local Acoustic Contribution in Speech LLMs via Word-Span Counterfactual Interventions
Related posts
Your RAG Finds Some of the Answer. I Tested 5 Chunking Strategies From This Month's Papers to See Who Finds All of It.
0dev.to
Goravel routing, route groups and controllers
0dev.to
Find and Replace in the Browser: Regex, Case Sensitivity and a Result You Can Check
0dev.to
I counted every subreddit since 2005, and "how many" depends on one definition
0dev.to
Remove EXIF Metadata in the Browser: How to Verify the Result Byte by Byte
0dev.to
I Built My Own Compiled Programming Language. Here's What I Learned.
0dev.to