Abstract Understanding the semantic meaning behind animal behavior remains a fundamental challenge in neuroscience and biomedical research. We present BehaVERT (BehaVior bERT), a transformer-based model that treats skeletal motion as structured language. By tokenizing movement into semantic units and learning temporal dependencies, BehaVERT decodes behavioral patterns from raw keypoint trajectories. Our pipeline addresses three critical gaps: accessibility, standardization, and interpretability. First, the pipeline provides web-based annotation tools for keypoint and behavior labeling that require no programming expertise, democratizing dataset creation. Second, this work contributes curated benchmarks with novel pose annotations and standardized formats, enabling reproducible evaluation across diverse behavioral tasks. Third, our transformer-based architecture processes temporal sequences to reveal latent meaning in movement, bridging the gap between immediate actions and long-term behavioral trends through both discrete behavior and continuous state classification. Novel training strategies including temporal unfolding and context-aware augmentation achieve state-of-the-art performance across five benchmarks: CalMS21, MABe22, PAIR-R24M, DeepEthogram, and SBeA. Beyond predictive accuracy, this work provides unprecedented interpretability: unsupervised clustering exposes learned behavioral structure, while attention analysis reveals which temporal moments and behaviors drive decisions. Our publicly accessible tools and analysis methods establish reproducible standards for decoding animal behavior.

With technology scaling down to 90nm and below, many yield-driven design and optimization methodologies have been proposed to cope with the prominent process variation and to increase the yield. A critical issue that affects the efficiency of those methods is to estimate the yield when given design parameters under variations. Existing methods either use Monte Carlo method in performance domain where thousands of simulations are required, or use local search in parameter domain where a number of simulations are required to characterize the point on the yield boundary defined by performance constraints. To improve efficiency, in this paper we propose QuickYield, a yield surface boundary determination by surface-point finding and global-search. Experiments on a number of different circuits show that for the same accuracy, QuickYield is up to 519X faster compared with the Monte Carlo approach, and up to 4.7X faster compared with YENSS, the fastest approach reported in literature.