Luis Garcés-Erice | Searcharxiv

arXiv2026

Towards Scalable Federated Container Orchestration: The CODECO Approach

This paper presents CODECO, a federated orchestration framework for Kubernetes that addresses the limitations of cloud-centric deployment. CODECO adopts a data-compute-network co-orchestration approach to support heterogeneous infrastructures, mobility, and multi-provider operation. CODECO extends Kubernetes with semantic application models, partition-based federation, and AI-assisted decision support, enabling context-aware placement and adaptive management of applications and their micro-services across federated environments. A hybrid governance model combines centralized policy enforcement with decentralized execution and learning to preserve global coherence while supporting far Edge autonomy. The paper describes the architecture and core components of CODECO, outlines representative orchestration workflows, and introduces a software-based experimentation framework for reproducible evaluation in federated Edge-Cloud infrastructure environments.

cs.DC↗

arXiv2025

Are Bias Evaluation Methods Biased ?

The creation of benchmarks to evaluate the safety of Large Language Models is one of the key activities within the trusted AI community. These benchmarks allow models to be compared for different aspects of safety such as toxicity, bias, harmful behavior etc. Independent benchmarks adopt different approaches with distinct data sets and evaluation methods. We investigate how robust such benchmarks are by using different approaches to rank a set of representative models for bias and compare how similar are the overall rankings. We show that different but widely used bias evaluations methods result in disparate model rankings. We conclude with recommendations for the community in the usage of such benchmarks.

cs.AI↗