Newest Papers
COVID-19 Exposures, Related Stressors, and the Health and Well-Being of U.S. Parents
Qi Li, Chris Knoester, Richard J. Petts, Daniel L. Carlson
Sociological Focus·2026
With the onset of COVID-19 in March 2020, U.S. parents faced new challenges in balancing work and family life while also navigating concerns about infection and disease prevention. In this study, we examine how exposure to COVID-19 and varying stressors during the pandemic shaped the health and well-being of U.S. parents. We use five waves of data collected between 2020 and 2023 from the Study on U.S. Parents’ Divisions of Labor during COVID-19. Growth curve model results show that COVID-19 infection per se does not significantly predict stress levels or depressive symptoms, although there is some evidence that recent COVID-19 infection is associated with declines in subjective health. But, work and family circumstances, both pre-pandemic and during the pandemic, apparently have a significant secondary impact on each health and well-being outcome. Finally, we find that the effects of certain work and family stressors vary by gender. These findings underscore the heterogeneous nature of pandemic experiences and highlight how the COVID-19 crisis may have shaped mothers’ and fathers’ health and well-being in distinct ways.
No ratings yet
View paper →What Does Flow Matching Bring To TD Learning?
Bhavya Agrawalla, Michal Nauman, Aviral Kumar
arXiv·2026
Recent work shows that flow matching can be effective for scalar Q-value function estimation in reinforcement learning (RL), but it remains unclear why or how this approach differs from standard critics. Contrary to conventional belief, we show that their success is not explained by distributional RL, as explicitly modeling return distributions can reduce performance. Instead, we argue that the use of integration for reading out values and dense velocity supervision at each step of this integration process for training improves TD learning via two mechanisms. First, it enables robust value prediction through \emph{test-time recovery}, whereby iterative computation through integration dampens errors in early value estimates as more integration steps are performed. This recovery mechanism is absent in monolithic critics. Second, supervising the velocity field at multiple interpolant values induces more \emph{plastic} feature learning within the network, allowing critics to represent non-stationary TD targets without discarding previously learned features or overfitting to individual TD targets encountered during training. We formalize these effects and validate them empirically, showing that flow-matching critics substantially outperform monolithic critics (2$\times$ in final performance and around 5$\times$ in sample efficiency) in settings where loss of plasticity poses a challenge e.g., in high-UTD online RL problems, while remaining stable during learning.
No ratings yet
View paper →