Saved in:
Bibliographic Details
Main Authors: Liu, Hui, Zou, Bin, Chen, Kecheng, Liu, Jie, Wang, Wenya, Li, Haoliang
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2604.09377
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915930250936320
author Liu, Hui
Zou, Bin
Chen, Kecheng
Liu, Jie
Wang, Wenya
Li, Haoliang
author_facet Liu, Hui
Zou, Bin
Chen, Kecheng
Liu, Jie
Wang, Wenya
Li, Haoliang
contents Large language models (LLMs) exhibit substantial variability in performance and computational cost across tasks and queries, motivating routing systems that select models to meet user-specific cost-performance trade-offs. However, existing routers generalize poorly in cold-start scenarios where in-domain training data is unavailable. We address this limitation with a multi-level task-profile-guided data synthesis framework that constructs a hierarchical task taxonomy and produces diverse question-answer pairs to approximate the test-time query distribution. Building on this, we introduce TRouter, a task-type-aware router approach that models query-conditioned cost and performance via latent task-type variables, with prior regularization derived from the synthesized task taxonomy. This design enhances TRouter's routing utility under both cold-start and in-domain settings. Across multiple benchmarks, we show that our synthesis framework alleviates cold-start issues and that TRouter delivers effective LLM routing.
format Preprint
id arxiv_https___arxiv_org_abs_2604_09377
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Task-Aware LLM Routing with Multi-Level Task-Profile-Guided Data Synthesis for Cold-Start Scenarios
Liu, Hui
Zou, Bin
Chen, Kecheng
Liu, Jie
Wang, Wenya
Li, Haoliang
Computation and Language
Large language models (LLMs) exhibit substantial variability in performance and computational cost across tasks and queries, motivating routing systems that select models to meet user-specific cost-performance trade-offs. However, existing routers generalize poorly in cold-start scenarios where in-domain training data is unavailable. We address this limitation with a multi-level task-profile-guided data synthesis framework that constructs a hierarchical task taxonomy and produces diverse question-answer pairs to approximate the test-time query distribution. Building on this, we introduce TRouter, a task-type-aware router approach that models query-conditioned cost and performance via latent task-type variables, with prior regularization derived from the synthesized task taxonomy. This design enhances TRouter's routing utility under both cold-start and in-domain settings. Across multiple benchmarks, we show that our synthesis framework alleviates cold-start issues and that TRouter delivers effective LLM routing.
title Task-Aware LLM Routing with Multi-Level Task-Profile-Guided Data Synthesis for Cold-Start Scenarios
topic Computation and Language
url https://arxiv.org/abs/2604.09377