Skip to main content
Search

University of South Florida

6 related articles

Subscribe to updates
  1. arXiv

    SinkState stores recurrent states as adaptive low-rank factors: hybrid-model decoding peak memory falls to as little as 1/2.6 of baseline

    Hybrid language models compress history into fixed-size recurrent states, yet serving systems still treat those states as dense matrices, reading and rewriting them whole at every decoding step, so memory traffic grows with batch size. SinkState exploits the concentration of queries, keys and values in linear attention into a stable low-dimensional subspace, keeping each state in an adaptive low-rank factorization throughout decoding, merging new tokens with a local SVD and consolidating the frozen part asynchronously on the CPU. Across four Mamba-2 and GDN hybrid models from 4B to 120B parameters, peak recurrent-state memory falls to 0.38–0.63 of the baseline and decoding speeds up by up to 1.58x, with accuracy maintained.

  2. Knee osteoarthritis grading without pooled records: ten sites hold feature columns, asynchronous federation reports 99.64% accuracy

    Knee osteoarthritis diagnosis draws on multimodal data scattered across institutions, and centralizing it is both costly and privacy-risky. An asynchronous federated scheme first fuses structured records, semi-structured files and imaging into a queryable data table, then has ten simulated institutions compute gradients on vertically partitioned feature columns, with Bulyan aggregation and adaptive differential privacy filtering malicious updates on the server. On 5,580 knee osteoarthritis cases it reports 99.64% accuracy, 0.998 AUC, 285 MB communication overhead and a 1.8% attack success rate.

  3. arXiv

    In MCP tool choice, the parameter schema decides: rename one field and selection drops from nine in ten to one in ten

    Among functionally similar MCP tools, the parameter schema shifts which tool a model picks more than names or descriptions do: renaming to as recipient_id cut selection of the same tool to 1–10% across five models.

  4. Statistica Sinica

    Delaunay-weighted two-sample test uses geometric direction information to detect principal-direction covariance differences in high-dimensional manifold data

    The authors propose a Delaunay-weighted two-sample test: under a low-dimensional manifold assumption they define a Delaunay weight from the Delaunay triangulation that captures both geodesic distance and relative direction, use the average within-group Delaunay weight as the test statistic with a permutation p-value, prove asymptotic normality under the null and consistency under the alternative, and show in simulations substantially higher power than k-NN, k-MST, kernel, e-distance, covariance, and regression tests when the two distributions differ in the principal directions of their covariance matrices, while detecting a treatment-group difference with p=0.011 in a mice protein expression dataset.

  5. Artificial Intelligence-Based Hypernasality Diagnosis Using CAPS-A-AM Rated Speech Samples in Pediatric Velopharyngeal Dysfunction

    In this prospective, single-center study, speech samples from 40 children aged 2 to 17 (including individuals with velopharyngeal dysfunction, conditions associated with VPD, and healthy participants) were collected during speech-language pathologist-guided evaluation with consensus CAPS-A-AM ratings, and mel spectrograms of high vowels /i/ and /u/ from sustained vowels, isolated words, and sentences were used to train logistic regression, an EfficientNet-V2-S attention multiple-instance-learning CNN, and a CNN-XGBoost hybrid for binary hypernasality classification at two CAPS-A-AM thresholds (absent 0 versus any hypernasality 1-4, and absent/borderline 0-1 versus mild-to-severe 2-4); multiple independent modeling approaches detected clinically rated hypernasality, with EfficientNet-V2-S a

  6. medRxiv

    An AI Competency Framework for Emergency Medicine: A Multiphase Consensus Process

    Using a Nominal Group Technique session (May 2025, n=6) and a two-round modified Delphi process (April–May 2026, expert panel n=13 across 12 academic medical centers, with 77% Round 1 and 100% Round 2 response rates), this study developed the first specialty-specific AI competency framework for a United States medical specialty, comprising 5 themes (Communicating about AI, Understanding appropriate use cases, Interacting with AI, AI risk management, Cognitive impacts of AI), 5 derived competencies (one-to-one theme-to-competency mapping endorsed by 12 of 13 panelists), 19 subthemes, and 10 retained clinical scenarios, with all themes, derived competencies, and individually rated subthemes meeting pre-specified consensus thresholds and 9 of 10 scenarios reaching consensus while 1 was retain

You’re all caught up