📄 中文摘要
现代大规模排名系统在复杂的竞争目标、操作约束和不断变化的产品需求中运行。该领域的进展越来越受到工程上下文约束的瓶颈影响，即将模糊的产品意图转化为合理、可执行和可验证的假设的艰难过程，而不仅仅是建模技术的限制。GEARS（生成性自主排名系统引擎）框架将排名优化重新构建为一个可编程实验环境中的自主发现过程。GEARS利用专业代理技能，将排名专家知识封装起来，超越了将优化视为静态模型选择的传统方法。
解码机器学习决策：大规模排名系统的自主推理框架
📄 English Summary
Decoding ML Decision: An Agentic Reasoning Framework for Large-Scale Ranking System
Modern large-scale ranking systems operate within a complex landscape of competing objectives, operational constraints, and evolving product requirements. Progress in this domain is increasingly bottlenecked by the engineering context constraint, which refers to the arduous process of translating ambiguous product intent into reasonable, executable, and verifiable hypotheses, rather than being limited by modeling techniques alone. The proposed framework, GEARS (Generative Engine for Agentic Ranking Systems), reframes ranking optimization as an autonomous discovery process within a programmable experimentation environment. By leveraging Specialized Agent Skills, GEARS encapsulates ranking expert knowledge, moving beyond the traditional approach of treating optimization as static model selection.