March 2026

Conventional federated learning (FL) frameworks often suffer from training degradation due to data uncertainty and heterogeneity across local clients. Probabilistic approaches such as Bayesian neural networks (BNNs) can mitigate this issue by explicitly modeling uncertainty, but they introduce additional runtime, latency, and bandwidth overhead that has rarely been studied in federated settings. To address these challenges, we propose Meta-BayFL, a personalized probabilistic FL method that combines meta-learning with BNNs to improve training under uncertain and heterogeneous data. The […]

Ver mais

Hi everyone, I recently published a paper on arXiv introducing a new ensemble learning framework called EARCP: https://arxiv.org/abs/2603.14651 EARCP is designed for sequential decision-making problems and dynamically combines multiple models based on both their performance and their agreement (coherence). Key ideas: Online adaptation of model weights using a multiplicative weights framework Coherence-aware regularization to stabilize ensemble behavior Sublinear regret guarantees: O(√(T log M)) Tested on time series forecasting, activity recognition, and financial prediction tasks The goal is to […]

Ver mais

Transformers enable in-context learning (ICL) for rapid, gradient-free adaptation in time series forecasting, yet most ICL-style approaches rely on tabularized, hand-crafted features, while end-to-end sequence models lack inference-time adaptation. We bridge this gap with a unified framework, Baguan-TS, which integrates the raw-sequence representation learning with ICL, instantiated by a 3D Transformer that attends jointly over temporal, variable, and context axes. To make this high-capacity model practical, we tackle two key hurdles: (i) calibration and training stability, improved with […]

Ver mais

Large Language Model (LLM) agents have shown strong results on multi-turn tool-use tasks, yet they operate in isolation during training, failing to leverage experiences accumulated across episodes. Existing experience-augmented methods address this by organizing trajectories into retrievable libraries, but they retrieve experiences only once based on the initial task description and hold them constant throughout the episode. In multi-turn settings where observations change at every step, this static retrieval becomes increasingly mismatched as episodes progress. We propose SLEA-RL […]

Ver mais

Following a joint demonstration at Embedded World (10–12 March), RemotiveLabs and Vayavya Labs announced a system-level ADAS validation setup that brings the full vehicle electronic architecture into software-in-the-loop (SIL). By combining Autoware, CARLA, and a virtual E/E architecture represented by RemotiveTopology, the companies demonstrated how OEMs and Tier 1s can validate autonomous driving behavior together with the vehicle’s electronic architecture – already at the SIL stage, before physical ECUs, HIL benches, or vehicle prototypes are available. The challenge: […]

Ver mais

This article presents a novel methodological framework for utilizing retrieval-augmented generation (RAG) systems specifically Google NotebookLM to address the critical challenge of tonal misalignment in competitive grant applications and institutional support documentation. By establishing a localized, source-grounded AI environment that synthesizes institutional data corpora with funder-specific linguistic patterns, this workflow enables rapid production of contextually precise, tonally calibrated grant narratives. The methodology operates in three phases: (1) source corpus ingestion and institutional knowledge mapping, (2) requirement-data alignment through […]

Ver mais

Zero-shot learning (ZSL) aims to recognize the unseen classes in the open-world guided by the side-information (e.g., attributes). Its key task is how to infer the latent semantic knowledge between visual and attribute features on seen classes, and thus conducting a desirable semantic knowledge transfer from seen classes to unseen ones. Prior works simply utilize unidirectional attention within a weakly-supervised manner to learn the spurious and limited latent semantic representations, which fail to effectively discover the intrinsic semantic […]

Ver mais

Every decentralized protocol built so far has shipped the same architectural mistake. Here’s what we borrowed from 150,000 years of human behavior to fix it. There’s a pattern in how anthropologists describe information flow in pre-literate communities. News doesn’t broadcast — it cascades. A person hears something from someone they trust, judges the claim against what they know of that source, and decides how far to pass it along. The further information travels from its origin, the more […]

Ver mais

Do you have many if/else conditions in your codebase? Or duplicated logic in both backend and UI? Does enabling a new set of similar behaviors mean a code change and a deployment? If so, it may be time to rethink the architecture. An alternative approach could be: data-driven architecture. Instead of encoding every nuance in code, we let data—configuration, schemas, and metadata—drive behavior. One configuration becomes the single source of truth for both backend and UI, allowing us […]

Ver mais