SkewRoute: Training-Free LLM Routing for Knowledge Graph Retrieval-Augmented Generation via Score Skewness of Retrieved Context
A training-free routing framework for knowledge graph retrieval-augmented generation.
The authors propose a simple and effective routing framework, SkewRoute, which balances performance and cost in knowledge graph retrieval-augmented generation. The framework is designed to direct queries to the most suitable language models based on score skewness of retrieved contexts. It achieves over 3x higher routing effectiveness while reducing runtime compared to existing methods.
Based on: SkewRoute: Training-Free LLM Routing for Knowledge Graph Retrieval-Augmented Generation via Score Skewness of Retrieved Context · arXiv (Cornell University)