成人影片

刘哲

讲座时间

2026年4月27日（周一）10:00

讲座地点

25教A315

01 成人影片 介绍

刘哲现任帝国理工成人影片商成人影片分析与运营系助理教授，研究方向为收益管理和供应链管理，侧重平台设计与分析、多源采购和库存控制。曾入围George Nicholson学生论文竞赛，获服务科学最佳论文一等奖、CSAMSE最佳论文二等奖、POMS及POMS-HK学生论文竞赛二等奖。现任系博士生项目主任、Decision Sciences副主编、Naval Research Logistics特邀副主编，并担任多项赛事与会议评委。刘哲本科毕业于清华大学工业工程系，获哥伦比亚大学商成人影片博士。

02 讲座内容

The rapid growth of Large Language Models (LLMs) has created a competitive ecosystem involving model providers (e.g., OpenAI, Anthropic) and a new layer of LLM aggregators (e.g., Poe, ChatHub), which offer unified access to multiple models. The aggregators exploit providers’ horizontal differentiation across tasks and vertical differentiation in performance to deliver a “best-of-both-worlds” service, generating additional wholesale API revenue for providers while cannibalizing their direct retail subscriptions. We develop a theoretical framework in which two differentiated LLM providers specialize in distinct features, and heterogeneous users choose between subscribing directly to a provider or to an aggregator. Our analysis shows that in early, unsaturated markets, aggregators act as market expanders by attracting users with balanced feature preferences, whereas in saturated markets, they cannibalize direct subscriptions unless providers’ API margins are sufficiently high. This reveals the providers’ tradeoff among subscription loss, API usage gain, and market expansion. Building on the static market segmentation analysis, we formulate a dynamic differential game and use a tractable marginal-revenue heuristic to analyze how providers strategically invest and innovate over time: improving a strong feature captures rival market share and boosts aggregator demand, while improving a weak feature reduces the aggregator’s appeal. Finally, incorporating heterogeneous usage intensity implies that light users seeking variety adopt the aggregator, while heavy users subscribe directly to the underlying LLM providers. This suggests tiered pricing or usage caps for the aggregators, as observed in practice.