用决策点自治约束大模型在推荐系统中的应用，平衡效率与可靠性。
RecSys Factory: Bounding LLM Agent Autonomy to Decision Points in the Industrial Recommender Lifecycle
- 只在关键决策点启用大模型，避免全流程失控
- 78天内成功执行1624次命令，成功率78.6%
- 适合工业级推荐系统中需要安全可控的团队
将大模型代理部署于工业推荐系统时面临三重矛盾：自主性（理解意图、零样本生成代码）、确定性（符合数据模式、不崩溃、无幻觉）和端到端效率。任选两者优化，必牺牲第三者。我们提出RecSys Factory平台，在腾讯三个异构推荐业务线中连续运行78天。核心设计为仅在决策点赋予自主权，而非贯穿整个流程。通过三项解构化解三重矛盾：运行时拆分为三种事件源（Claude Code Stop钩子、企业IM webhook、工作流调度API），等待阶段无长时进程，94%的等待时间零CPU消耗；能力拆分为29个文件的技能生态（8971行SKILL.md），每项技能的潜在陷阱自动编译成400条记录的PitfallStore，将自主性限制在预设流程的有界类型决策面内；部署覆盖三类业务，涵盖不同标签语义、A/B测试拓扑和操作员角色。其中两个业务线在上线初期出现显著加速，但未做对照实验，仅为观察现象。人类保留在诊断与执行边界，采用人机协同卡协议作为审计日志，具备结构化验证、幂等性和可重放性，8天16轮试点中有效运行。78天内共记录1624次命令调用，综合成功率78.6%。
原文摘要 · Abstract (English)
Deploying LLM agents into industrial recommender operations exposes a three-way tension we frame as the autonomy-determinism-efficiency trilemma: general autonomy (interpreting operator intent, generating glue code zero-shot), industrial determinism (schema-conforming feature extraction, non-crashing A/B, zero compliance-path hallucination), and end-to-end efficiency. Any two can be maximized against the third. We present RecSys Factory, an LLM-agent platform deployed for 78 days across three heterogeneous Tencent recommender business lines. The design principle is autonomy at decision points, not over pipelines, made concrete through three deconstructions that each discharge one vertex of the trilemma. Runtime is deconstructed into three host-emitted event sources (Claude Code Stop hooks, corporate-IM webhooks, workflow scheduler APIs): the platform carries no long-running daemon during the wait phase and consumes zero CPU during the 94% of wall-clock spent waiting on Spark or GPU jobs. Capability is deconstructed into a 29-file skill ecosystem (8,971 lines of SKILL.md) whose per-skill pitfall tables mechanically compile into a 400-entry PitfallStore, confining autonomy to bounded typed decision surfaces inside pre-committed pipelines. Deployment spans three business lines with disjoint label semantics, A/B layer topologies, and operator personas; an onboarding-time compression is observed on two of the three and is reported as a case-study observation, not a generalization claim, and not measured against a controlled pre-platform baseline. The human is retained at the diagnostic-versus-execution boundary via a human-in-the-loop card protocol, deployed as an audit-trail primitive (schema-validated, idempotent, replayable) and reported from an 8-day 16-run pilot. Across the 78-day window the platform recorded 1,624 CLI-tool dispatches at a 78.6% aggregate success rate.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。