Routing Should Pay for Itself: Sparse Supervision for Economical LLM Routing

2026-09-29

AI Summary

This study addresses the high supervision costs and overlooked upfront investment recovery in LLM routing by proposing SaveRouter, a framework that first quantifies routing economic benefits and reveals the oversupply inherent in dense supervision. To reduce training overhead, it introduces sparse supervision with selective feedback acquisition, combined with cross-query capability information sharing and fine-grained routing optimization. Experimental results demonstrate that only 33%–41% of feedback is sufficient to maintain high-quality routing, reducing the break-even point by 1.9× to 9.5×. By achieving joint optimization of supervision expenditure and service savings, SaveRouter significantly enhances the overall cost-effectiveness of routing systems.

Abstract

Large language model (LLM) routing reduces serving cost by assigning each query to an appropriate model while preserving response quality. Learning such a router, however, often requires executing multiple candidate models on historical queries to collect query--model quality feedback, creating a nontrivial supervision cost before deployment. Existing work largely focuses on serving-time efficiency, overlooking whether the resulting savings are sufficient to recover this upfront expenditure. We further observe that routing quality often saturates well before all query--model feedback is collected, suggesting that dense supervision can be economically over-provisioned. We propose SaveRouter, a sparse-supervision routing framework that selectively acquires informative model feedback and shares capability information across related queries, while retaining query-level refinement for fine-grained routing. We evaluate routing by jointly accounting for supervision expenditure and subsequent serving-time savings. Across four routing benchmarks, the main setting uses only about 33--41% of available training feedback while maintaining competitive or better routing quality, and reduces the break-even deployment volume by approximately 1.9--9.5 times compared with the fastest conventional router. Further analysis shows that acquiring more supervision is not always economically preferable: the supervision level that minimizes serving cost can differ from the one that achieves the earliest payback. Our code is publicly available at https://github.com/LAMDA-Model-Reuse/SaveRouter.

Problem

• LLM routing
• supervision cost
• cost efficiency
• sparse supervision
• break-even analysis

Innovation

• Sparse Supervision
• LLM Routing
• Cost Efficiency
• SaveRouter
• Break-even Analysis