Research & Practice
Research Intern | Kuaishou Technology
March 2023 - Present
Supervisor: Qingpeng Cai
Research
研究推荐系统可控性,尝试解耦 LLM 功能,构建特殊的工作流增加用户对 LLM 推荐系统的可控性。
Practice
在精排阶段,对 PXTR 与 ensemble score 的逆序对数量为 loss,进行单步帕累托,充分发挥 PXTR 的打分信息。
Impact & Achievements
Production Metrics:
Average app time: +0.120% | Follow rate: +1.002% | Forward rate: +0.984% | Download rate: +1.018% | Collection rate: +1.102% | Negative feedback: -2.627%
Research Output:
Published 1 paper at WSDM 2025 based on work conducted during the internship.
