Bioacoustics, the study of sounds produced by living organisms, plays a vital role in conservation, biodiversity monitoring, and behavioral studies. Many tasks in this field, such as species, individual, and behavior classification and detection, are well-suited to machine learning. However, they often suffer from limited annotated data, highlighting the need for a general-purpose bioacoustic encoder capable of extracting useful representations for diverse downstream tasks. Such encoders have been proposed before, but are often limited in scope due to a focus on a narrow range of species (typically birds), and a reliance on a single model architecture or training paradigm. Moreover, they are usually evaluated on a small set of tasks and datasets. In this work, we present a large-scale empirical study that covers aspects of bioacoustics that are relevant to research but have previously been scarcely considered: training data diversity and scale, model architectures and training recipes, and the breadth of evaluation tasks and datasets. We obtain encoders that are state-of-the-art on the existing and proposed benchmarks. We also identify what matters for training these encoders, such that this work can be extended when more data are available or better architectures are proposed. Specifically, across 26 datasets with tasks including species classification, detection, individual ID, and vocal repertoire discovery, we find self-supervised pre-training followed by supervised post-training on a mixed bioacoustics + general-audio corpus yields the strongest in- and out-of-distribution performance. We show the importance of data diversity in both stages. To support ongoing research and application, we will release the model checkpoints.

While language models are increasingly utilized in materials science, typical models rely on frequency-centric tokenization methods originally developed for natural language processing. However, these methods frequently produce excessive fragmentation and semantic loss, failing to maintain the structural and semantic integrity of material concepts. To address this issue, we propose MATTER, a novel tokenization approach that integrates material knowledge into tokenization. Based on MatDetector trained on our materials knowledge base and a re-ranking method prioritizing material concepts in token merging, MATTER maintains the structural integrity of identified material concepts and prevents fragmentation during tokenization, ensuring their semantic meaning remains intact. The experimental results demonstrate that MATTER outperforms existing tokenization methods, achieving an average performance gain of $4\%$ and $2\%$ in the generation and classification tasks, respectively. These results underscore the importance of domain knowledge for tokenization strategies in scientific text processing. Our code is available at https://github.com/yerimoh/MATTER

About a year ago, Frankfurt's Lord Mayor Peter Feldmann had the impulse to launch an AI initiative. Stefan Jäger, a speaker in the Lord Mayor's office and honorary board member of the association, let us know what Feldmann had in mind: "As always with new technologies, citizens have difficulty imagining what artificial intelligence actually is in this case. The association wants to acquire and share knowledge. Only transparency will make people curious." And Dr Thorsten Pötter wants to help him do so.

Monge matrices and their permuted versions known as pre-Monge matrices naturally appear in many domains across science and engineering. While the rich structural properties of such matrices have long been leveraged for algorithmic purposes, little is known about their impact on statistical estimation. In this work, we propose to view this structure as a shape constraint and study the problem of estimating a Monge matrix subject to additive random noise. More specifically, we establish the minimax rates of estimation of Monge and pre-Monge matrices. In the case of pre-Monge matrices, the minimax-optimal least-squares estimator is not efficiently computable, and we propose two efficient estimators and establish their rates of convergence. Our theoretical findings are supported by numerical experiments.

Long before computers did in fact conquer the world, the influential German electronic group Kraftwerk calculated that probability, contemplating the changes to come by harnessing the latest musical gear and technology to create enduring work that soundtracked the birth of the Digital Age. On Sunday night, the quartet made its Hollywood Bowl debut by offering an overview of its career, including the metronomic 1974 classic "Autobahn," the menacing antinuke song "Radio-Activity," the genre-defining electro classic "Trans-Europe Express," the important synthesizer pop gems from the group's album "Computer World" and more. Mesmerizing to experience in the open air, the band looped bloops and beeps, thumps and bumps and sibilant fake high-hats to play songs once described by Detroit techno producer Carl Craig as "so stiff they were funky." The show was accompanied by 3-D visuals that locked image with music and required the 17,000-odd fans to don glasses, making the crowd seem teleported from an Atomic Age movie theater. The sight and sound presented a persuasive argument on the band's enduring influence, even if it revealed the ways that, like the beige and boxy early-era personal computers, the march of time renders even the most innovative technological expressions obsolete.