arXiv:2501.10077v2 Announce Type: replace-cross Abstract: The double descent phenomenon challenges traditional statistical learning theory by revealing scenarios where larger models do not necessarily lead to reduced performance on unseen data. While this counterintuitive behavior has been observed in a variety…
arXiv:2509.15932v1 Announce Type: cross Abstract: Large language models improve with scale, yet feedback-based alignment still exhibits systematic deviations from intended behavior. Motivated by bounded rationality in economics and cognitive science, we view judgment as resource-limited and feedback as a constrained…
arXiv:2501.18164v4 Announce Type: replace-cross Abstract: We theoretically analyzed the convergence behavior of Riemannian stochastic gradient descent (RSGD) and found that using an increasing batch size leads to faster convergence than using a constant batch size, not only with a constant…
arXiv:2509.15981v1 Announce Type: cross Abstract: In reinforcement learning with sparse rewards, demonstrations can accelerate learning, but determining when to imitate them remains challenging. We propose Smooth Policy Regularisation from Demonstrations (SPReD), a framework that addresses the fundamental question: when should…
arXiv:2509.15593v1 Announce Type: new Abstract: In transfer learning, a source domain often carries diverse knowledge, and different domains usually emphasize different types of knowledge. Different from handling only a single type of knowledge from all domains in traditional transfer learning…
The Universal Tool Calling Protocol (UTCP) is a lightweight, secure, and scalable way for AI agents and applications to find and call tools directly, without the need for additional wrapper servers. Key Features The Problem with Current Approaches Traditional solutions…
Meta researchers introduced a method that compresses repeated reasoning patterns into short, named procedures—“behaviors”—and then conditions models to use them at inference or distills them via fine-tuning. The result: up to 46% fewer reasoning tokens on MATH while matching or…
While it seems like a foregone conclusion that there will be a foldable iPhone, possibly late next year, there hasn’t been much info about what it would look like. In the latest installment of his Power On newsletter Mark Gurman…
Analyst Ming-Chi Kuo took to X on Wednesday to claim that a MacBook Pro with an OLED touchscreen was expected to enter mass production by late 2026. Today Bloomberg’s Mark Gurman is following up to remind us that he was…
