Federate the Router: Learning Language Model Routers with Sparse and Decentralized Evaluations
This addresses the challenge of optimizing LLM routing for edge and enterprise clients with decentralized data, though it is incremental as it builds on existing routing methods with a federated approach.
The paper tackles the problem of routing queries to large language models (LLMs) with fragmented and privacy-sensitive evaluation data across clients, introducing a federated framework that improves the accuracy-cost frontier by increasing effective model coverage and better query generalization across two benchmarks.
Large language models (LLMs) are increasingly accessed as remotely hosted services by edge and enterprise clients that cannot run frontier models locally. Since models vary widely in capability and price, routing queries to models that balance quality and inference cost is essential. Existing router approaches assume access to centralized query-model evaluation data. However, these data are often fragmented across clients, such as end users and organizations, and are privacy-sensitive, which makes centralizing data infeasible. Additionally, per-client router training is ineffective since local evaluation data is limited and covers only a restricted query distribution and a biased subset of model evaluations. We introduce the first federated framework for LLM routing, enabling clients to learn a shared routing policy from local offline query-model evaluation data. Our framework supports both parametric multilayer perceptron router and nonparametric K-means router under heterogeneous client query distributions and non-uniform model coverage. Across two benchmarks, federated collaboration improves the accuracy-cost frontier over client-local routers, both via increased effective model coverage and better query generalization. Our theoretical results also validate that federated training reduces routing suboptimality.