digitado

We propose a new partial-observability model for online learning problems where the learner, besides its own loss, also observes some noisy feedback about the other actions, depending on the underlying structure of the problem. We represent this structure by a weighted directed graph, where the edge weights are related to the quality of the feedback shared by the connected nodes. Our main contribution is an efficient algorithm that guarantees a regret of $widetilde{O}(sqrt{α^* T})$ after $T$ rounds, where […]

Ver mais

Google is changing the way we think about browsers. Instead of just being a place to access websites, Chrome is trying to become more interactive by introducing “Skills.” The company has made it clear that it wants AI to be part of your everyday browsing, not just something you use in separate chat windows. It will help you create structured, reusable instructions that act more like shortcuts than conversations. At its core, this feature makes Chrome even more […]

Ver mais

Reinforcement learning (RL) enables high-frequency, closed-loop control for robotic manipulation, but scaling to long-horizon tasks with sparse or imperfect rewards remains difficult due to inefficient exploration and poor credit assignment. Vision-Language-Action (VLA) models leverage large-scale multimodal pretraining to provide generalist, task-level reasoning, but current limitations hinder their direct use in fast and precise manipulation. In this paper, we propose Vision-Language-Action Jump-Starting (VLAJS), a method that bridges sparse VLA guidance with on-policy RL to improve exploration and learning efficiency. […]

Ver mais

Let $ xi(z) $ be the Riemann xi function. In a previous paper we prove the boundedness of coefficients of the power series expansion of $ xi'(1/z)/xi(1/z) $ and thus give a proof of the Riemann hypothesis. In this paper we generalize the method there to the study of the extended Riemann hypothesis for general number fields.

Ver mais

Network access control and identity legitimacy verification have been implemented by establishing a secure foundation for the trusted establishment of communication entities. However, successful identity authentication alone does not guarantee secure communication. In open-network environments, it remains essential to establish a secure session key via a robust key agreement mechanism—one that prevents explicit disclosure of identity information while ensuring post-quantum security. To address these requirements, we propose a lattice-based key agreement protocol. The protocol integrates identity binding, implicit […]

Ver mais

The classical binary heap sink operation based on swap has a significant write overhead. We examine two intuitive improvements: swapping siblings (verified via bounded SMT search) and adding a local hint called pref (the hint-assisted variant). In our bounded SMT checks and implementation comparisons, we did not find evidence that these variants provide consistent benefits; PerfView measurements show the hint-assisted variant was slower in most configurations. Our results suggest that reverting to the straightforward hole-based sink is the […]

Ver mais

This study addresses the problem of zero-shot generalization (ZSG) in deep reinforcement learning by proposing an MNK game strategy learning method based on a Fully Convolutional Deep Q-Network (FCN-DQN). Research in deep reinforcement learning aims to develop algorithms that can generalize well to unseen environments at deployment time, thereby avoiding overfitting to the training environment. Solving this problem is crucial for real-world applications, where environments are diverse, dynamic, and inherently unpredictable. By constructing a fully convolutional reinforcement learning […]

Ver mais