arXiv cs.CL· Jihao Zhao, Ding Chen, Yanfang Chen, Bo Tang, Feiyu Xiong, Junchi Yan, Zhiyu Li·· 11 小时前AI 评分31
RouteAlign：按任务需求对齐模型画像的 LLM 服务选择方法
RouteAlign: Aligning Task Requirements with Model Profiles for LLM Service Selection
AI 导读
RouteAlign 通过任务需求刻画能力需求与难度，并将其与模型性能测量对齐，用于 LLM 服务选择。它用历史需求构建全局画像与局部能力估计，再由共享打分器整合证据、以生成成本偏好筛选候选服务。在多任务与异构 LLM 服务评测中，该方法将质量-成本前沿推向更高质量，相比开源 Jev 风格决策后端质量稳定提升且生成成本更低。
正文 · 原文
Abstract:LLM service selection requires performance evidence that reflects the demands of the current task. Aggregate scores obscure differences in capability and difficulty, while semantically related examples can involve different combinations of skills. We present RouteAlign, which uses task requirements to structure both the measurement of model performance and the selection of services. Task requirements describe capability demand and difficulty, enabling alignment with corresponding model performance measurements. Historical requirements determine how outcomes contribute to global profiles and local capability estimates; incoming requirements determine how those measurements inform the current decision. A shared scorer integrates the evidence, and a generation-cost preference controls selection among eligible services. Evaluation across diverse tasks and heterogeneous LLM services shows that RouteAlign extends the empirical quality-cost frontier toward higher task quality. Comparisons with open Jev-style decision backends show consistent quality gains, with lower generation cost than their strongest tested configuration. Ablations support the contribution of local performance evidence, while catalog-update experiments demonstrate effective reuse of the fitted scorer with refreshed model measurements.
| Subjects: | Software Engineering (cs.SE); Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.08162 [cs.SE] |
| (or arXiv:2611.08162v1 [cs.SE] for this version) | |
| https://doi.org/10.48550/arXiv.2611.08162 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Zhiyu Li [view email]
[v1]
Thu, 5 Nov 2026 15:21:16 UTC (1,308 KB)
来源：arXiv cs.CL · arxiv.org