智能的影子价格：作为供应链问题的大语言模型推理质量退化
The Shadow Price of Intelligence: Quality Degradation in LLM Inference as a Supply Chain Problem
浏览论文内容
中文总结 AI 辅助
本文将LLM推理质量退化视为供应链问题，建模拥塞下的服务分配，揭示节流的反直觉影响，提出影子价格可毫秒级计算最优策略，证明拥塞时节流是需求杠杆而非成本杠杆。
中文摘要 AI 辅助
大语言模型（LLM）提供商受计算资源约束，面对拥塞时普遍采取降低服务质量的应对措施：将查询路由到更小的模型、减少推理工作量、截断上下文。行业核算认为此举可节省成本，但本文证明该核算有误，因为其在客户购买答案时对查询定价，而质量下降的答案会以一定概率失效，失效的答案要么作为重试返回，在系统负载最高时增加到达量，要么作为流失离开，在无成本仪表盘显示的账本上破坏终身价值。本文用三个经典原语对推理分配建模：报童模型，其缺货成本为流失的终身价值；几何重试乘数，其中回收的产品是不满；双体制瞬态队列，其到达率因重试而内生。静态来看，存在非空可测体制，在此体制中，更便宜的模型每获得一个满意答案可节省能量，但每获得一个满意答案会消耗更多容量，因此当容量受限时，折扣会完全反转。动态来看，激增期间触发的反应式节流可能越过点火阈值，在此阈值之上，其产生的流量比减少的更多，而设置在退化均衡以下的释放规则会将瞬态激增转化为永久退化体制。在异构客户情况下，节流是重试膨胀负载中的运输问题，其最优策略按临界比率、逐类配给智能，其对偶即智能的影子价格，按类和小时对边际查询定价，闭式轨迹使其可在毫秒内计算。随机分析强化而非削弱了该论点：点火边界获得预测宽度，噪声会惩罚将系统停在边界上的反应式策略。在拥塞情况下，节流不是成本杠杆，而是需求杠杆。
英文摘要
Large language model providers are compute constrained, and a common response to congestion is to degrade service. A degraded answer fails with some probability, and a failed answer either returns as a retry or departs as churn, destroying LTV on an unaccounted ledger. We model inference allocation as a newsvendor whose stockout cost is churned lifetime value, a geometric retry multiplier in which the recycled product is dissatisfaction, and a two-regime transient fluid queue whose arrival rate is made endogenous by retries. Statically, there is a regime in which a cheaper model saves energy per initiated task while consuming strictly more capacity per initiated task, so the discount inverts when capacity binds. Dynamically, a reactive throttle fired during a surge can cross an ignition threshold beyond which it manufactures more traffic than it sheds, and a release rule set below the degraded equilibrium converts a transient surge into a permanent degraded regime. With heterogeneous customers, throttling is a transportation problem in retry-inflated load whose optimal policy rations intelligence by critical ratio, and whose dual, the shadow price of intelligence, prices a marginal query by class and by hour. Under congestion, throttling is not a cost lever but a demand lever.
发表机构
- Lehigh University - College of Business（里海大学商学院）
机构由 AI 辅助整理，请以论文原文为准。
