cs.CL提交于 2026-10-15已译

RouteWeaver: 将模式选择与执行编织进统一的大语言模型路由

RouteWeaver: Weaving Mode Selection and Execution into Unified LLM Routing

Xiaohan Wang · Haozhen Zhang · Qingyuan Liu · Tao Feng · Wenya Wang

中文摘要

大语言模型（LLM）路由在多样化任务中动态选择和协调异构语言模型。现有方法通常基于预定义的执行范式（如单轮、多轮或智能体路由），固定了模型调用组织方式，限制了针对不同查询需求的适应性。RouteWeaver 提出一种统一路由框架，在单一 LLM 策略中整合这些路由模式，联合学习使用哪种模式以及在该模式下如何选择模型。为联合优化模式选择和模式内执行，开发了 COMET-GRPO（协调模式-执行训练），将轨迹级优势解耦为模式选择和模式内执行的独立信号，实现对每个决策层级的更有针对性的信用分配。此外引入渐进式探索，训练从外部指定模式逐步过渡到自主选择，使路由器先学会如何执行每种模式，再决定何时使用。在数学、编码、知识、逻辑推理和事实回忆五个领域评估，RouteWeaver 平均准确率为 61.4%，相对最强基线提升 11.03%，并在所有五个领域超越所有对比基线。消融实验表明 COMET-GRPO 和渐进式探索对自适应路由均至关重要；若无渐进式探索，路由策略完全退化为单轮执行。

关键要点

01现有 LLM 路由依赖预定义执行范式，固定模型调用组织方式，难以适应不同查询需求。
02RouteWeaver 在单一 LLM 策略中联合学习路由模式选择与模式内模型选择，实现统一路由。
03COMET-GRPO 将轨迹级优势解耦为模式选择与模式内执行信号，实现更精准的信用分配。
04渐进式探索使路由器先掌握各模式执行方式，再学习自主模式选择，避免策略退化。
05在数学、编码等五个领域平均准确率 61.4%，相对最强基线提升 11.03%，全面超越对比基线。

解读

尚无解读。

原始英文摘要

arXiv:2610.19705v1 Announce Type: new Abstract: LLM routing dynamically selects and coordinates heterogeneous language models across diverse tasks. However, existing approaches are largely built around a predefined execution paradigm, such as single-round, multi-round, or agentic routing, which fixes how model calls are organized and limits adaptation to different query needs. We propose RouteWeaver, a unified routing framework that integrates these routing modes within a single LLM policy, jointly learning which mode to use and how to select models within that mode. To jointly optimize mode selection and within-mode execution, we develop COMET-GRPO (COordinated Mode-Execution Training), which decouples the trajectory-level advantage into separate signals for mode selection and within-mode execution, enabling more targeted credit assignment to each decision level. We further introduce progressive exploration, which gradually shifts training from externally assigned modes to autonomous selection, so that the router first learns how to execute each mode before deciding when to use it. We evaluate RouteWeaver across mathematics, coding, knowledge, logical reasoning, and factual recall. RouteWeaver achieves an average accuracy of 61.4%, an 11.03% relative improvement over the strongest baseline, and outperforms all compared baselines in each of the five domains. Ablations show that both COMET-GRPO and progressive exploration are important for adaptive routing; without progressive exploration, the routing policy collapses entirely to single-round execution.