Ahmed Metwally

Sirui Li, Shuhan Xiao, Mihir Joshi et al.

The rise of large language models (LLMs) has shifted time series analysis from narrow analytics to general-purpose reasoning. Yet, existing benchmarks cover only a small set of health time series modalities and tasks, failing to reflect the diverse domains and extensive temporal dependencies inherent in real-world physiological modeling. To bridge these gaps, we introduce HEARTS (Health Reasoning over Time Series), a unified benchmark for evaluating hierarchical reasoning capabilities of LLMs over general health time series. HEARTS integrates 16 real-world datasets across 12 health domains and 20 signal modalities, and defines a comprehensive taxonomy of 110 tasks grouped into four core capabilities: Perception, Inference, Generation, and Deduction. Evaluating 14 state-of-the-art LLMs on more than 20K test samples reveals intriguing findings. First, LLMs substantially underperform specialized models, and their performance is only weakly related to general reasoning scores. Moreover, LLMs often rely on simple heuristics and struggle with multi-step temporal reasoning. Finally, performance declines with increasing temporal complexity, with similar failure modes within model families, indicating that scaling alone is insufficient. By making these gaps measurable, HEARTS provides a standardized testbed and living benchmark for developing next-generation LLM agents capable of reasoning over diverse health signals.

Maxwell A. Xu, Girish Narayanswamy, Kumar Ayush et al.

Foundation models, a cornerstone of recent advancements in machine learning, have predominantly thrived on complete and well-structured data. Wearable sensor data frequently suffers from significant missingness, posing a substantial challenge for self-supervised learning (SSL) models that typically assume complete data inputs. This paper introduces the second generation of Large Sensor Model (LSM-2) with Adaptive and Inherited Masking (AIM), a novel SSL approach that learns robust representations directly from incomplete data without requiring explicit imputation. AIM's core novelty lies in its use of learnable mask tokens to model both existing ("inherited") and artificially introduced missingness, enabling it to robustly handle fragmented real-world data during inference. Pre-trained on an extensive dataset of 40M hours of day-long multimodal sensor data, our LSM-2 with AIM achieves the best performance across a diverse range of tasks, including classification, regression and generative modeling. Furthermore, LSM-2 with AIM exhibits superior scaling performance, and critically, maintains high performance even under targeted missingness scenarios, reflecting clinically coherent patterns, such as the diagnostic value of nighttime biosignals for hypertension prediction. This makes AIM a more reliable choice for real-world wearable data applications.

Ahmed Metwally, Chun-Heng Huang

This work tackles the problem of fuzzy joining of strings that naturally tokenize into meaningful substrings, e.g., full names. Tokenized-string joins have several established applications in the context of data integration and cleaning. This work is primarily motivated by fraud detection, where attackers slightly modify tokenized strings, e.g., names on accounts, to create numerous identities that she can use to defraud service providers, e.g., Google, and LinkedIn. To detect such attacks, all the accounts are pair-wise compared, and the resulting similar accounts are considered suspicious and are further investigated. Comparing the tokenized-string features of a large number of accounts requires an intuitive tokenized-string distance that can detect subtle edits introduced by an adversary, and a very scalable algorithm. This is not achievable by existing distance measure that are unintuitive, hard to tune, and whose join algorithms are serial and hence unscalable. We define a novel intuitive distance measure between tokenized strings, Normalized Setwise Levenshtein Distance (NSLD). To the best of our knowledge, NSLD is the first metric proposed for comparing tokenized strings. We propose a scalable distributed framework, Tokenized-String Joiner (TSJ), that adopts existing scalable string-join algorithms as building blocks to perform NSLD-joins. We carefully engineer optimizations and approximations that dramatically improve the efficiency of TSJ. The effectiveness of the TSJ framework is evident from the evaluation conducted on tens of millions of tokenized-string names from Google accounts. The superiority of the tokenized-string-specific TSJ framework over the general-purpose metric-spaces joining algorithms has been established.