# RACER: Risk-Aware Calibrated Efficient Routing for Large Language Models

Source: [RACER: Risk-Aware Calibrated Efficient Routing for Large Language Models](<https://www.sota2.com/research/paper/69afa57c2d603e6de3e439b4>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-03-11T05:00:04.181Z

Authors: Sai Hao, Hao Zeng, Hongxin Wei, Bingyi Jing

Published: 2026-02-20T08:23:03.000Z

## Abstract

Efficiently routing queries to the optimal large language model (LLM) is crucial for optimizing the cost-performance trade-off in multi-model systems. However, most existing routers rely on single-model selection, making them susceptible to misrouting. In this work, we formulate LLM routing as the $\\alpha$-VOR problem to minimize expected set size while controlling the misrouting risk, and propose a novel method -- RACER, extending base routers to output model sets that can be subsequently aggregated for improved output. In particular, RACER constructs nested model sets via augmented scoring and utilizes finite-sample concentration bounds to calibrate a threshold that allows for both variable set sizes and abstention. We theoretically prove that RACER achieves rigorous distribution-free risk control on unseen test data in a post-hoc and model-agnostic manner. Extensive experiments verify our theoretical guarantees and demonstrate that RACER consistently enhances downstream accuracy across a wide range of benchmarks.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Multitask Language Understanding on C-MMLU (Accuracy)](<https://www.sota2.com/research/sota/multitask-language-understanding-on-c-mmlu-accuracy.md>): Multitask Language Understanding; C-MMLU; Accuracy (C-MMLU): 50.7.

- [Multitask Language Understanding on MMLU (Accuracy)](<https://www.sota2.com/research/sota/multitask-language-understanding-on-mmlu-accuracy.md>): Multitask Language Understanding; MMLU; Accuracy: 63.7.

- [Question Answering on ARC Challenge](<https://www.sota2.com/research/sota/question-answering-on-arc-challenge.md>): Question Answering; ARC Challenge; Accuracy: 56.8.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2603.06616>)

[PDF](<https://arxiv.org/pdf/2603.06616.pdf>)