Background: Digital phenotyping uses passively collected smartphone-sensing data to characterize everyday behavior in naturalistic settings, and has become an important approach for studying mental well-being. Most previous studies have examined associations between individual sensing variables and mental health. However, mental well-being is likely reflected not by isolated behaviors but by combinations of co-occurring daily behaviors that together form lifestyles. Person-centered approaches capable of identifying these behavioral configurations may, therefore, provide more interpretable digital phenotypes; yet, such approaches have rarely been applied to passive smartphone-sensing data. Objective: This study aimed to examine whether smartphone-captured behavioral and environmental data could be used to derive interpretable day-level and person-level lifestyle profiles, and whether person-level profiles were associated with mental well-being. We also tested whether Big Five personality traits—extraversion, agreeableness, conscientiousness, openness, and negative emotionality—moderated these associations. Methods: The study used a 2-week prospective longitudinal cohort design with a sample of 553 German adults (mean age 42.12, SD 12.89 years; 44.65% female) drawn from an initial sample recruited according to quotas designed to reflect the German population. Ten smartphone-sensing indicators captured 5 domains, including communication and social media app use, mobility, physical activity, environmental context, and phone-use intensity. Mental well-being was assessed using the Warwick–Edinburgh Mental Well-Being Scale, and personality was assessed using the 15-item Big Five Inventory–2 Extra-Short Form. We used multilevel latent profile analysis to identify day-level profiles nested within person-level profiles. Associations between profiles and mental well-being were tested using classification-error–adjusted mean comparisons and omnibus Wald tests. Moderation was examined using hierarchical regressions comparing models with and without profile-by-personality interactions. Results: Eight day-level profiles and 7 person-level profiles were identified. Day-level profiles reflected distinct combinations of smartphone-sensing indicators. Person-level profiles represented different distributions of these daily patterns. Profiles differed significantly only in positive functioning (Wald ²=13.39; =.04), not in overall mental well-being, positive affect, or satisfying interpersonal relationships. The physically active and unplugged profile had higher positive functioning than the mobile and always-on social profile (mean 3.94, SD 0.63 vs mean 3.61, SD 0.74; Cohen =0.47; 95% CI 0.21‐0.73). No other pairwise differences were significant. Sensitivity analyses excluding the smallest profile produced comparable results, supporting the robustness of the findings. Personality-by-profile interactions did not significantly improve prediction for any well-being outcome. Conclusions: The findings extend the field by showing that transparent, person-centered digital phenotypes can distinguish variation in positive functioning, although causal conclusions cannot be drawn. In real-world settings, such interpretable profiles could support understandable monitoring tools and, following prospective replication and validation, inform personalized multibehavior interventions that target combinations of behaviors rather than single behaviors in isolation.

Nature, Published online: 08 April 2026; doi:10.1038/s41586-026-10260-w

Daily satellite data reveal that Earth’s artificial lights at night are highly volatile, with frequent brightening and dimming between 2014 and 2022.

arXiv:2603.17445v3 Announce Type: replace Abstract: When a multi-agent system produces an incorrect or harmful answer, who is accountable if execution logs and agent identifiers are unavailable? In practice, generated content is often detached from its execution environment due to privacy or system boundaries, leaving the final text as the only auditable artifact. Existing attribution methods rely on full execution traces and thus become ineffective in such metadata-deprived settings. We propose Implicit Execution Tracing (IET), a provenance-by-design framework that shifts attribution from post-hoc inference to built-in instrumentation. Instead of reconstructing hidden trajectories, IET embeds agent-specific, key-conditioned statistical signals directly into the token generation process, transforming the output text into a self-verifying execution record. At inference time, we recover a linearized execution trace from the final text via transition-aware statistical scoring. Experiments across diverse multi-agent coordination settings demonstrate that IET achieves accurate segment-level attribution and reliable transition recovery under identity removal, boundary corruption, and privacy-preserving redaction, while maintaining generation quality. These results show that embedding provenance into generation provides a practical and robust foundation for accountability in multi-agent language systems when execution metadata is unavailable.

arXiv:2509.25106v3 Announce Type: replace-cross Abstract: Deep Research Agents (DRAs) can autonomously conduct complex investigations and generate comprehensive reports, demonstrating strong real-world potential. However, existing evaluations mostly rely on close-ended benchmarks, while open-ended deep research benchmarks remain scarce and typically neglect personalized scenarios. To bridge this gap, we introduce Personalized Deep Research Bench (PDR-Bench), the first benchmark for evaluating personalization in DRAs. It pairs 50 diverse research tasks across 10 domains with 25 authentic user profiles that combine structured persona attributes with dynamic real-world contexts, yielding 250 realistic user-task queries. To assess system performance, we propose the PQR Evaluation Framework, which jointly measures Personalization Alignment, Content Quality, and Factual Reliability. Our experiments on a range of systems highlight current capabilities and limitations in handling personalized deep research. This work establishes a rigorous foundation for developing and evaluating the next generation of truly personalized AI research assistants.