Oldest Papers
Estimating the impact of out-of-home placement on health risk behavior in adolescents exposed to maltreatment: An advanced causal inference approach
Austin J. Blake, Mariola Moeyaert, Felix J. Thoemmes, David Mackinnon, Laurie Chassin
Child Abuse & Neglect·2025
Background
Youth who experience out-of-home placement (OOHP) engage in elevated health risk behaviors (e.g., substance use, unprotected sex), with risk potentially heightened for those placed during adolescence. Estimating causal effects is challenging because maltreated youth who are placed differ systematically from those who remain in-home.
Objective
This study examined the effects of adolescent OOHP on health risk behaviors, applying causal inference methods (g-estimation and inverse probability of treatment weighting; IPTW) to address selection bias and time-varying confounders.
Participants and setting
Data were drawn from 734 maltreated adolescents in the U.S. National Survey on Child and Adolescent Wellbeing.
Methods
IPTW and g-estimation were used to estimate effects of adolescent OOHP on substance use and sexual risk behavior during adolescence and into young adulthood, adjusting for numerous confounders. Results were compared with regression analyses using traditional covariate adjustment.
Results
In traditional regression models, OOHP was not significantly associated with health risk behaviors. However, both causal inference approaches revealed that OOHP predicted increased substance use later in adolescence. IPTW analyses also indicated greater sexual risk behavior in adolescence and increased substance use in adulthood among placed youth.
Conclusions
Although OOHP is intended to enhance safety, adolescent OOHP may heighten risk for harmful health behaviors. Given the severe consequences of such behaviors and the risk of losing service access in adulthood, the period following OOHP is a critical window for intervention. The results demonstrate how robust causal inference techniques may lead to more accurate assessment of OOHP than traditional regression methods.
No ratings yet
View paper →Deepseek, ChatGPT, and Gemini versus human subtitling: a case study in socio-cultural adaptation in multimedia communication
Peng Cao, Masood Khoshsaligheh, Fatemeh Jomhouri
Perspectives·2025·1 citations
No abstract available
No ratings yet
View paper →What AI Might Mean For Workers: A Discussion - American Compass
Oren Cass
American Compass·2025
No abstract available
No ratings yet
View paper →Thinking Augmented Pre-training
Liang Wang, Nan Yang, Shaohan Huang, Li Dong, Furu Wei
arXiv·2025
This paper introduces a simple and scalable approach to improve the data efficiency of large language model (LLM) training by augmenting existing text data with thinking trajectories. The compute for pre-training LLMs has been growing at an unprecedented rate, while the availability of high-quality data remains limited. Consequently, maximizing the utility of available data constitutes a significant research challenge. A primary impediment is that certain high-quality tokens are difficult to learn given a fixed model capacity, as the underlying rationale for a single token can be exceptionally complex and deep. To address this issue, we propose Thinking augmented Pre-Training (TPT), a universal methodology that augments text with automatically generated thinking trajectories. Such augmentation effectively increases the volume of the training data and makes high-quality tokens more learnable through step-by-step reasoning and decomposition. We apply TPT across diverse training configurations up to $100$B tokens, encompassing pre-training with both constrained and abundant data, as well as mid-training from strong open-source checkpoints. Experimental results indicate that our method substantially improves the performance of LLMs across various model sizes and families. Notably, TPT enhances the data efficiency of LLM pre-training by a factor of $3$. For a $3$B parameter model, it improves the post-training performance by over $10\%$ on several challenging reasoning benchmarks.
No ratings yet
View paper →