Your personalresearch library

Discover papers, track what you're reading, rate and review.Build a library that grows with your curiosity.

Get Started Explore Papers

Currently Reading

Want to Read

All Papers

Advanced filtersTopic: All

Experimental

Topic

First Proof

Mohammed Abouzaid, Andrew J. Blumberg, Martin Hairer, Joe Kileel, Tamara G. Kolda, Paul D. Nelson, Daniel Spielman, Nikhil Srivastava, Rachel Ward, Shmuel Weinberger, Lauren Williams

arXiv·2026

To assess the ability of current AI systems to correctly answer research-level mathematics questions, we share a set of ten math questions which have arisen naturally in the research process of the authors. The questions had not been shared publicly until now; the answers are known to the authors of the questions but will remain encrypted for a short time.

No ratings yet

View paper →

Daily Health Problems, Views of Aging, and the Moderating Role of Awareness of Age-Related Changes

Maiken Tingvold, Shevaun D Neupert, Yuval Palgi, Amit Shrira, Anna E Kornadt

The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences·2026

Abstract Objectives Views of aging, such as subjective age or subjective accelerated aging, are related to health problems: In longitudinal and daily assessments, experiencing more health issues is associated with more negative views of aging. This study investigates whether the association between health problems and multiple views of aging constructs is moderated by people’s experience of age-related gains and losses. We therefore expected people to demonstrate more negative views of aging on days with more health problems. Following previous research on awareness of age-related changes as an important moderator for the impact of age-related experiences to developmental outcomes, we assumed that 1) on days when participants experienced higher awareness of age-related gains the adverse effect of health problems on views of aging would be reduced, 2) the association between health problems and views of aging should be amplified on days when participants experienced higher awareness of age-related losses. Methods A sample of N = 69 participants aged 52-75 years (M age = 62.72, SD = 5.57) reported their subjective age (uni- and multidimensional), subjective accelerated aging, health problems, and awareness of age-related gains and losses for up to 14 days of daily diary assessments. Age, gender, education, and baseline health were included as covariates. Results Multilevel models showed that perceiving more age-related losses was associated with an exacerbation of the positive association between daily health problems and multidimensional subjective age and subjective accelerated aging. Discussion Our findings underscore the importance of perceiving age-related losses in daily life. Perceiving changes as age-related may influence how daily experiences are interpreted and their impact on developmental outcomes.

No ratings yet

View paper →

PreviousPage 24 of 610Next

Rethinking the Trust Region in LLM Reinforcement Learning

Penghui Qi, Xiangxin Zhou, Zichen Liu, Tianyu Pang, Chao Du, Min Lin, Wee Sun Lee

arXiv·2026

Reinforcement learning (RL) has become a cornerstone for fine-tuning Large Language Models (LLMs), with Proximal Policy Optimization (PPO) serving as the de facto standard algorithm. Despite its ubiquity, we argue that the core ratio clipping mechanism in PPO is structurally ill-suited for the large vocabularies inherent to LLMs. PPO constrains policy updates based on the probability ratio of sampled tokens, which serves as a noisy single-sample Monte Carlo estimate of the true policy divergence. This creates a sub-optimal learning dynamic: updates to low-probability tokens are aggressively over-penalized, while potentially catastrophic shifts in high-probability tokens are under-constrained, leading to training inefficiency and instability. To address this, we propose Divergence Proximal Policy Optimization (DPPO), which substitutes heuristic clipping with a more principled constraint based on a direct estimate of policy divergence (e.g., Total Variation or KL). To avoid huge memory footprint, we introduce the efficient Binary and Top-K approximations to capture the essential divergence with negligible overhead. Extensive empirical evaluations demonstrate that DPPO achieves superior training stability and efficiency compared to existing methods, offering a more robust foundation for RL-based LLM fine-tuning.

No ratings yet

View paper →

Generative Modeling via Drifting

Mingyang Deng, He Li, Tianhong Li, Yilun Du, Kaiming He

arXiv·2026

Generative modeling can be formulated as learning a mapping f such that its pushforward distribution matches the data distribution. The pushforward behavior can be carried out iteratively at inference time, for example in diffusion and flow-based models. In this paper, we propose a new paradigm called Drifting Models, which evolve the pushforward distribution during training and naturally admit one-step inference. We introduce a drifting field that governs the sample movement and achieves equilibrium when the distributions match. This leads to a training objective that allows the neural network optimizer to evolve the distribution. In experiments, our one-step generator achieves state-of-the-art results on ImageNet at 256 x 256 resolution, with an FID of 1.54 in latent space and 1.61 in pixel space. We hope that our work opens up new opportunities for high-quality one-step generation.

No ratings yet

View paper →

Expert Selections In MoE Models Reveal (Almost) As Much As Text

Amir Nuriyev, Gabriel Kulp

arXiv·2026

We present a text-reconstruction attack on mixture-of-experts (MoE) language models that recovers tokens from expert selections alone. In MoE models, each token is routed to a subset of expert subnetworks; we show these routing decisions leak substantially more information than previously understood. Prior work using logistic regression achieves limited reconstruction; we show that a 3-layer MLP improves this to 63.1% top-1 accuracy, and that a transformer-based sequence decoder recovers 91.2% of tokens top-1 (94.8% top-10) on 32-token sequences from OpenWebText after training on 100M tokens. These results connect MoE routing to the broader literature on embedding inversion. We outline practical leakage scenarios (e.g., distributed inference and side channels) and show that adding noise reduces but does not eliminate reconstruction. Our findings suggest that expert selections in MoE deployments should be treated as sensitive as the underlying text.

No ratings yet

View paper →

Your personalresearch library

All Papers

First Proof

Daily Health Problems, Views of Aging, and the Moderating Role of Awareness of Age-Related Changes

Parental Gender Identity Socialization Among Transgender and Gender Expansive Youth in the U.S.: A Grounded Theory Conceptual Model

On Optimism for Interpretability

You and Your Research Agent: Lessons From Using Agents for Interpretability Research

Intentionally Designing the Future of AI

Discrimination, depressive symptoms, and prescription opioid misuse among adults with chronic pain who engage in hazardous drinking

Is Tradition a “Curse”? — Reflections on Jujutsu Kaisen

Rethinking the Trust Region in LLM Reinforcement Learning

Generative Modeling via Drifting

How Jeff Bezos Brought Down the Washington Post | The New Yorker

Expert Selections In MoE Models Reveal (Almost) As Much As Text