Newest papers

Learning to Refer: Inference-Time Collaboration in Decentralized Federated Learning

Yuheng Ma, Chendi Wang, Qiong Zhang · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26800 · cited by 0 works (OpenAlex) · open-access PDF

Spanish

En el aprendizaje federado descentralizado de datos, los clientes heterogéneos pueden desarrollar competencias complementarias, pero cada consulta suele ser evaluada sólo por el cliente donde llega. Para aprovechar esta experiencia distribuida sin coordinación global, formulamos la colaboración en tiempo de inferencia como referencias de expertos locales presupuestadas, un proceso de enrutamiento limitado por gráficos en el que una consulta solo puede moverse entre clientes vecinos y solo los clientes visitados contribuyen a su predicción final. EdgeDRM implementa este marco al aprender un potencial de enrutamiento basado en densidad para cada cliente: las diferencias entre potenciales vecinos proporcionan las comparaciones utilizadas para seleccionar referencias sucesivas, mientras que los modelos de tareas locales generan predicciones que se agregan a lo largo de la ruta resultante. Tanto la formación como la inferencia son descentralizadas, utilizando

Machine translation (Meta M2M100); the English original follows.

Abstract

In decentralized federated learning, heterogeneous clients may develop complementary expertise, yet each query is typically evaluated only by the client where it arrives. To exploit this distributed expertise without global coordination, we formulate inference-time collaboration as budgeted local expert referral, a graph-constrained routing process in which a query can move only between neighboring clients and only visited clients contribute to its final prediction. EdgeDRM implements this framework by learning a density-based routing potential for each client: differences between neighboring potentials provide the comparisons used to select successive referrals, while local task models generate predictions that are aggregated along the resulting route. Both training and inference are decentralized, using only neighbor-local information without a central router or graph-wide client scoring. We establish estimation guarantees for the learned comparisons and characterize how local improvement and graph structure govern prior-weighted density gains from referral. Experiments on benchmark and real-world medical data show that EdgeDRM outperforms graph-matched decentralized baselines in most settings, with the clearest benefits under sparse connectivity.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Exploit, Inference, Task (project management), Process (computing), Benchmark (surveying)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange