Residual reinforcement learning (RL) enables stable online refinement of expressive pretrained policies by freezing the base and learning only bounded corrections. However, value learning in residual RL poses unique challenges that remain poorly understood. In this work, we identify two key bottlenecks: cold start pathology, where the critic lacks knowledge of the value landscape around the base policy, and structural scale mismatch, where the residual contribution is dwarfed by the base action. Through systematic investigation, we uncover the […]

Ver mais

Machine learning models can represent climate processes that are nonlocal in horizontal space, height, and time, often by combining information across these dimensions in highly nonlinear ways. While this can improve predictive skill, it makes learned relationships difficult to interpret and prone to overfitting as the extent of nonlocal information grows. We address this challenge by introducing data-driven integration kernels, a framework that adds structure to nonlocal operator learning by explicitly separating nonlocal information aggregation from local nonlinear […]

Ver mais

arXiv:2601.09887v1 Announce Type: new Abstract: Contemporary materials science research is heavily conducted in silico, involving massive simulations of the atomic-scale evolution of materials. Cataloging basic patterns in the atomic displacements is key to understanding and predicting the evolution of physical properties. However, the combinatorial complexity of the space of possible transitions coupled with the overwhelming amount of data being produced by high-throughput simulations make such an analysis extremely challenging and time-consuming for domain experts. The development of visual […]

Ver mais

arXiv:2604.06187v1 Announce Type: new Abstract: Artificial intelligence is increasingly used in hiring, raising concerns about how applicants perceive these systems. While prior work on algorithmic fairness has emphasized technical bias mitigation, little is known about how avatar identity cues influence applicants’ justice attributions in an interview context. We conducted a crowdsourcing study with 215 participants who completed an interview with photorealistic AI avatars varied in phenotypic traits (race and sex), followed by a standardized rejection. Using self-reports, sentiment […]

Ver mais

arXiv:2602.02582v1 Announce Type: new Abstract: Large language models (LLMs) enable powerful zero-shot recommendations by leveraging broad contextual knowledge, yet predictive uncertainty and embedded biases threaten reliability and fairness. This paper studies how uncertainty and fairness evaluations affect the accuracy, consistency, and trustworthiness of LLM-generated recommendations. We introduce a benchmark of curated metrics and a dataset annotated for eight demographic attributes (31 categorical values) across two domains: movies and music. Through in-depth case studies, we quantify predictive uncertainty (via […]

Ver mais

On January 16, 2026, John Lee Dumas of Entrepreneurs on Fire hosted the 2026 Podcast Hall of Fame ceremony, which was a relatively intimate affair, unlike the elaborate red carpet events for movie stars and entertainers. A cash bar and appetizer station occupied one corner of the venue, nothing extravagant, but enough to make the space feel welcoming before the program began. The evening balanced live induction speeches with pre-recorded remarks from inductees who could not attend in person. The remote […]

Ver mais

Engineering teams are experiencing one of the fastest shifts in tooling since the rise of modern IDEs and cloud development. AI coding software tools have moved far beyond autocomplete—developers can now describe intent in natural language and watch fully formed implementations appear. This creates a new foundation for productivity, creativity, and learning across engineering organizations. For many teams, this rapid evolution is redefining how engineers work, moving from manual, line-by-line implementation toward higher-level collaboration with AI. Teams are now […]

Ver mais

All prior membership inference attacks for fine-tuned language models use hand-crafted heuristics (e.g., loss thresholding, Min-K%, reference calibration), each bounded by the designer’s intuition. We introduce the first transferable learned attack, enabled by the observation that fine-tuning any model on any corpus yields unlimited labeled data, since membership is known by construction. This removes the shadow model bottleneck and brings membership inference into the deep learning era: learning what matters rather than designing it, with generalization through training […]

Ver mais

For most of the internet era, educational technology followed a familiar pattern: digitize the textbook, put worksheets on a screen, maybe add a video. The goal was efficiency. Faster distribution. Easier grading. More content, more quickly. But that model is starting to feel dated. Today’s learners don’t just want information delivered. They expect to interact with it. Remix it. Build on top of it. The question is no longer whether technology belongs in education; it’s whether our education […]

Ver mais

arXiv:2604.13065v1 Announce Type: new Abstract: LLMs can execute every step of chain-of-thought reasoning correctly and still produce wrong final answers. We introduce the Novel Operator Test, a benchmark that separates operator logic from operator name, enabling rigorous distinction between genuine reasoning and pattern retrieval. By evaluating Boolean operators under unfamiliar names across depths 1-10 on five models (up to 8,100 problems each), we demonstrate a reasoning-output dissociation that existing benchmarks cannot detect. At Claude Sonnet 4’s depth 7, […]

Ver mais