从采样结果到能力分布：重新思考LLM路由的监督
From Sampled Outcomes to Capability Distributions: Rethinking Supervision for LLM Routing
AI总结：
针对LLM路由中单次响应作为监督信号噪声大的问题，提出DARS框架，从分布视角构建路由监督，考虑输入和输出不确定性，实验表明分布感知监督更稳定有效。
摘要：
现有的LLM路由方法通常将模型对查询的单个响应作为训练路由器的能力标签。然而，由于LLM生成本质上是随机的，这种单次监督仅提供了查询-模型对行为的噪声观测，而非可靠的能力估计。我们表明，这种假设会向路由监督中引入系统性噪声，使得学习到的路由策略可靠性降低。为解决此问题，我们提出DARS（分布感知路由监督）框架，该框架从模型行为的分布视角构建路由监督。DARS不依赖单个生成的响应，而是考虑来自输入侧和输出侧的不确定性，捕捉语义等价的查询表述和随机生成如何影响模型性能。基于这些分布感知的观测，DARS为路由构建更可靠的监督信号。跨不同任务的实验表明，单次标签可能对模型选择产生误导，而分布感知监督提供更稳定的标签并改进学习到的路由行为。我们的结果表明，可靠的LLM路由应超越单次响应观测，并基于查询级模型能力分布。
英文摘要：
Existing LLM routing methods often construct supervision from a single sampled response for each query--model pair. Because LLM generation is stochastic, however, such an observation can be an unstable estimate of model capability: semantically equivalent query formulations and repeated decoding may yield different scores and even different model preferences. We show that this instability can further propagate from routing labels to learned routing policies. To address this issue, we propose DARS (Distribution-Aware Routing Supervision), which estimates query-level model capability from repeated observations spanning semantics-preserving query rewrites and stochastic decoding. DARS summarizes expected quality, expected cost, and performance variability to construct risk-aware supervision without changing the downstream router architecture. Experiments across diverse tasks and routing methods show that DARS generally improves routing utility and cost--quality trade-offs over single-shot supervision. Further analyses show that its benefits persist under moderate sampling budgets and different decoding temperatures. These results suggest that reliable LLM routing should move beyond individual sampled outcomes and instead model query-level capability distributions.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
