RouteMoA: Dynamic Routing without Pre-Inference Boosts Efficient Mixture-of-Agents
This addresses efficiency problems for users of large language model ensembles, though it is incremental as it builds on existing MoA methods.
The paper tackles the high cost and latency of Mixture-of-Agents (MoA) by proposing RouteMoA, a framework that uses dynamic routing to reduce inference needs, achieving an 89.8% cost reduction and 63.6% latency reduction in large-scale model pools.
Mixture-of-Agents (MoA) improves LLM performance through layered collaboration, but its dense topology raises costs and latency. Existing methods employ LLM judges to filter responses, yet still require all models to perform inference before judging, failing to cut costs effectively. They also lack model selection criteria and struggle with large model pools, where full inference is costly and can exceed context limits. To address this, we propose RouteMoA, an efficient mixture-of-agents framework with dynamic routing. It employs a lightweight scorer to perform initial screening by predicting coarse-grained performance from the query, narrowing candidates to a high-potential subset without inference. A mixture of judges then refines these scores through lightweight self- and cross-assessment based on existing model outputs, providing posterior correction without additional inference. Finally, a model ranking mechanism selects models by balancing performance, cost, and latency. RouteMoA outperforms MoA across varying tasks and model pool sizes, reducing cost by 89.8% and latency by 63.6% in the large-scale model pool.