RADAR: Reasoning-Ability and Difficulty-Aware Routing for Reasoning LLMs
This work addresses the practical deployment challenge of balancing cost and performance for reasoning LLMs, offering an incremental improvement in routing efficiency.
The paper tackles the tradeoff between performance and cost in deploying reasoning language models by introducing RADAR, a routing framework that assigns queries to model-budget pairs based on difficulty and ability, achieving superior performance on 8 reasoning benchmarks and showing generalization to out-of-distribution queries.
Reasoning language models have demonstrated remarkable performance on many challenging tasks in math, science, and coding. Choosing the right reasoning model for practical deployment involves a performance and cost tradeoff at two key levels: model size and reasoning budget, where larger models and higher reasoning budget lead to better performance but with increased cost and latency. In this work, we tackle this tradeoff from the angle of model configuration routing for different queries, and present RADAR (Reasoning-Ability and Difficulty-Aware Routing), a lightweight, interpretable, and scalable routing framework. Inspired by psychometrics, RADAR learns an item response model from model responses with different budgets to different queries, with interpretable parameters including query difficulties and model-budget abilities. RADAR then routes queries with higher difficulty to model-budget pairs with higher ability, and vice versa. We conduct extensive experiments on 8 widely used challenging reasoning benchmarks, demonstrating the superior performance of RADAR compared to state-of-the-art model routing methods. RADAR also exhibits query generalization capabilities, showing strong performance on out-of-distribution queries in all benchmarks. RADAR is also scalable and can efficiently integrate additional models by dynamically selecting a small set of evaluation queries to estimate their abilities.