digitado

The deployment of deep neural networks in safety-critical systems necessitates reliable and efficient uncertainty quantification (UQ). A practical and widespread strategy for UQ is repurposing stochastic regularizers as scalable approximate Bayesian inference methods, such as Monte Carlo Dropout (MCD) and MC-DropBlock (MCDB). However, this paradigm remains under-explored for Stochastic Depth (SD), a regularizer integral to the residual-based backbones of most modern architectures. While prior work demonstrated its empirical promise for segmentation, a formal theoretical connection to Bayesian variational […]

Ver mais

Autonomous AI agents operating in high-stakes domains—financial trading, medical diagnostics, autonomous code execution—lack formal safety guarantees for their core operational loops, including memory management, tool invocations, and human interactions. Current verification approaches either fail to scale to neural components or ignore the structured control flow of agentic systems entirely. We introduce AgentVerify (Compositional Formal Verification of AI Agent Safety Properties via LTL Model Checking), a model checking framework that specifies and verifies safety properties for agent architectures using […]

Ver mais

Recent advances in deep learning underscore the need for systems that can not only acquire new knowledge through Continual Learning (CL) but also remove outdated, sensitive, or private information through Machine Unlearning (MU). However, while CL methods are well-developed, MU techniques remain in early stages, creating a critical gap for unified frameworks that depend on both capabilities. We find that naively combining existing CL and MU approaches results in knowledge leakage a gradual degradation of foundational knowledge across […]

Ver mais

The relevance of this study stems from the need for a scientifically sound assessment of the environmental risks associated with launch vehicle launches and for ensuring the environmental safety of areas potentially impacted by space activities. Comprehensive environmental monitoring in the impact areas of rocket parts and adjacent populated areas is particularly important, taking into account natural and climatic factors and the spatial heterogeneity of pollution. This study assessed the environmental impacts of the “Soyuz-2.1a” launch with the […]

Ver mais

Text-based web agents offer computational efficiency for autonomous web navigation, yet developing robust agents remains challenging due to the noisy and heterogeneous nature of real-world HTML. Standard Supervised Fine-Tuning (SFT) approaches fail in two critical dimensions: they lack discrimination capabilities to reject plausible but incorrect elements in densely populated pages, and exhibit limited generalization to unseen website layouts. To address these challenges, we introduce the Triton dataset (590k instances) and a progressive training curriculum. Triton is constructed via […]

Ver mais

We introduce KumoRFM-2, the next iteration of a pre-trained foundation model for relational data. KumoRFM-2 supports in-context learning as well as fine-tuning and is applicable to a wide range of predictive tasks. In contrast to tabular foundation models, KumoRFM-2 natively operates on relational data, processing one or more connected tables simultaneously without manual table flattening or target variable generation, all while preserving temporal consistency. KumoRFM-2 leverages a large corpus of synthetic and real-world data to pre-train across four […]

Ver mais