HuggingFace Daily Papers·· 2 天前
TokenRouter：面向 Token 级 LLM 路由的高效服务系统
TokenRouter: Efficient Serving System for Token-Level LLM Routing
AI 导读
TokenRouter 是一个面向 Token 级 LLM 路由的高效服务系统，遵循「请求中心编程、模型中心执行」原则，为每个 LLM 启动子服务器并异步调度请求，其延迟批处理调度器超参数由吞吐数学模型推导。在多种路由算法、负载和模型组合下，TokenRouter 的解码吞吐量比现有系统高 2.01-64.15 倍。该工作已被 NeurIPS 2026 接收，代码已开源。
正文
Abstract:Large language model (LLM) routing distributes inference work across different models, advancing the cost-quality Pareto frontier of LLM serving. While coarse-grained routing at the session or query level has been widely adopted in production systems, recent algorithmic work shows that fine-grained token-level routing can yield substantial efficiency and quality gains. However, efficiently serving token-level routed inference poses significant challenges to existing systems. Built on single-LLM assumptions, current systems suffer from severe step desynchronization and frequent batch admission delays under token-level routing, and they also impose high implementation complexity on developers. To address these challenges, we design TokenRouter, an efficient and developer-friendly serving system for token-level routed LLM inference. TokenRouter follows the principle of request-centric programming, model-centric execution: developers describe routing logic from the perspective of a single request, while the runtime launches a subserver for each LLM and dispatches requests asynchronously. Each subserver employs a delayed-batching scheduler, whose optimal hyperparameters are derived from a mathematical throughput model of the system. Across diverse routing algorithms, workloads, and model pairs, TokenRouter achieves 2.01-64.15x higher decoding throughput than existing systems, substantially advancing the serving efficiency of token-level LLM routing. Our code is available at this https URL.
| Comments: | Accepted by NeurIPS 2026 |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.12242 [cs.CL] |
| (or arXiv:2610.12242v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.12242 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Tengxuan Liu [view email]
[v1]
Thu, 8 Oct 2026 16:21:50 UTC (329 KB)
来源：HuggingFace Daily Papers · arxiv.org