从负反馈中学习并衡量序列推荐系统的响应能力

thbcm阅读(134)

在工业级推荐系统中,尤其是短视频与信息流场景,用户行为以
隐式反馈为主。点击、长时观看通常被视为正反馈,而大量未点击、快速滑过、短停留(例如小于 3 秒)的行为,长期被简单地视为“负样本”或被忽略。

关于Agentic AI的一些总结和思考

thbcm阅读(128)

Agentic AI frameworks are software toolkits that provide structure, components, and protocols for building autonomous AI agents that can reason, plan, and act to achieve goals, streamlining development with features like memory, tool integration (APIs), and multi-agent orchestration, with popular examples including 
AutoGenCrewAI, LangGraph, and Semantic Kernel, enabling complex automation beyond simple prompts. 

联系我们