Your personalresearch library

We critically examine the limitations of current AI models in achieving autonomous learning and propose a learning architecture inspired by human and animal cognition. The proposed framework integrates learning from observation (System A) and learning from active behavior (System B) while flexibly switching between these learning modes as a function of internally generated meta-control signals (System M). We discuss how this could be built by taking inspiration on how organisms adapt to real-world, dynamic environments across evolutionary and developmental timescales.

No ratings yet

View paper →

Differences in Assessing Loneliness Among Japanese Older Adults: A Comparison of Family Physicians and Nurses

Kazutaka Yoshida, Aya Goto, Ichiro Kawachi

Journal of Clinical Medicine·2026

Background/Objectives: Loneliness is highly prevalent in Japan and has become a major public health concern. Although primary health care professionals are often the first to encounter lonely patients, loneliness is subjective and difficult to detect in routine clinical practice. This study aims to examine how family physicians and nurses assess patient loneliness, and whether their approaches differ. Methods: This mixed-methods study comprised two surveys administered in Japanese family medicine clinics. Survey 1 (August 2020) was a cross-sectional questionnaire involving patients aged ≥ 50 years (n = 470), six family physicians, and seven nurses, of whom one responded on behalf of the group. Patient loneliness was measured using the UCLA (University of California, Los Angeles) Loneliness Scale (Version 3) and served as the reference standard. Physicians and nurses independently assessed patient loneliness based on medical records. Sensitivity, specificity, and predictive values were calculated. Survey 2 (August–September 2023) used an open-ended questionnaire completed by the same physicians and nurses, with responses analyzed using quantitative text mining to explore assessment perspectives. Results: Based on the UCLA scale, 38% of patients were classified as lonely. Compared to each other, family physicians demonstrated a higher sensitivity (45.3%) but lower specificity (67.4%), whereas nurses showed a lower sensitivity (21.8%) but higher specificity (84.5%). Text mining revealed that family physicians emphasized relational quality and psychological context, sometimes identifying loneliness even when there is no apparent lack of social connections. In contrast, nurses tended to define loneliness in terms of clearly observable social circumstances and emphasized patients’ subjective acknowledgment. Conclusions: Family physicians and nurses employ distinct yet complementary approaches to identifying loneliness in primary health care. Collaborative, role-based strategies may enhance the accurate detection of loneliness and support more effective patient care.

No ratings yet

View paper →

Page 1 of 554Next

Reject, Resample, Repeat: Understanding Parallel Reasoning in Language Model Inference

Noah Golowich, Fan Chen, Dhruv Rohatgi, Raghav Singhal, Carles Domingo-Enrich, Dylan J. Foster, Akshay Krishnamurthy

arXiv·2026

Inference-time methods that aggregate and prune multiple samples have emerged as a powerful paradigm for steering large language models, yet we lack any principled understanding of their accuracy-cost tradeoffs. In this paper, we introduce a route to rigorously study such approaches using the lens of *particle filtering* algorithms such as Sequential Monte Carlo (SMC). Given a base language model and a *process reward model* estimating expected terminal rewards, we ask: *how accurately can we sample from a target distribution given some number of process reward evaluations?* Theoretically, we identify (1) simple criteria enabling non-asymptotic guarantees for SMC; (2) algorithmic improvements to SMC; and (3) a fundamental limit faced by all particle filtering methods. Empirically, we demonstrate that our theoretical criteria effectively govern the *sampling error* of SMC, though not necessarily its final *accuracy*, suggesting that theoretical perspectives beyond sampling may be necessary.

No ratings yet

View paper →

A recipe for scalable attention-based MLIPs: unlocking long-range accuracy with all-to-all node attention

Eric Qu, Brandon M. Wood, Aditi S. Krishnapriyan, Zachary W. Ulissi

arXiv·2026

Machine-learning interatomic potentials (MLIPs) have advanced rapidly, with many top models relying on strong physics-based inductive biases. However, as models scale to larger systems like biomolecules and electrolytes, they struggle to accurately capture long-range (LR) interactions, leading current approaches to rely on explicit physics-based terms or components. In this work, we propose AllScAIP, a straightforward, attention-based, and energy-conserving MLIP model that scales to O(100 million) training samples. It addresses the long-range challenge using an all-to-all node attention component that is data-driven. Extensive ablations reveal that in low-data/small-model regimes, inductive biases improve sample efficiency. However, as data and model size scale, these benefits diminish or even reverse, while all-to-all attention remains critical for capturing LR interactions. Our model achieves state-of-the-art energy/force accuracy on molecular systems, as well as a number of physics-based evaluations (OMol25), while being competitive on materials (OMat24) and catalysts (OC20). Furthermore, it enables stable, long-timescale MD simulations that accurately recover experimental observables, including density and heat of vaporization predictions.

No ratings yet

View paper →

Causal Interpretation of Neural Network Computations with Contribution Decomposition

Joshua Brendan Melander, Zaki Alaoui, Shenghua Liu, Surya Ganguli, Stephen A. Baccus

arXiv·2026

Understanding how neural networks transform inputs into outputs is crucial for interpreting and manipulating their behavior. Most existing approaches analyze internal representations by identifying hidden-layer activation patterns correlated with human-interpretable concepts. Here we take a direct approach to examine how hidden neurons act to drive network outputs. We introduce CODEC (Contribution Decomposition), a method that uses sparse autoencoders to decompose network behavior into sparse motifs of hidden-neuron contributions, revealing causal processes that cannot be determined by analyzing activations alone. Applying CODEC to benchmark image-classification networks, we find that contributions grow in sparsity and dimensionality across layers and, unexpectedly, that they progressively decorrelate positive and negative effects on network outputs. We further show that decomposing contributions into sparse modes enables greater control and interpretation of intermediate layers, supporting both causal manipulations of network output and human-interpretable visualizations of distinct image components that combine to drive that output. Finally, by analyzing state-of-the-art models of neural activity in the vertebrate retina, we demonstrate that CODEC uncovers combinatorial actions of model interneurons and identifies the sources of dynamic receptive fields. Overall, CODEC provides a rich and interpretable framework for understanding how nonlinear computations evolve across hierarchical layers, establishing contribution modes as an informative unit of analysis for mechanistic insights into artificial neural networks.

No ratings yet

View paper →

Your personalresearch library

All Papers

Health and well-being after spousal loss among older men and women

Perceived Antisemitism and Mental Health in Jewish Mothers and Their Children

The Incredible Story of the Cartel Olympics - The Atlantic

Taming the Trapezius - Vitae Aether

Perceived discrimination and employment anxiety among impoverished students in Chinese higher vocational colleges: roles of relative deprivation and rumination

Why AI systems don't learn and what to do about it: Lessons on autonomous learning from cognitive science

Differences in Assessing Loneliness Among Japanese Older Adults: A Comparison of Family Physicians and Nurses

Cisheterosexism, minority stress, and mental health challenges among LGBTQ+ adults in Spain

The Westernization Of Moaning

Reject, Resample, Repeat: Understanding Parallel Reasoning in Language Model Inference

A recipe for scalable attention-based MLIPs: unlocking long-range accuracy with all-to-all node attention

Causal Interpretation of Neural Network Computations with Contribution Decomposition