DuoLM：基于语言模型的高效个性化推荐框架
原标题：DuoLM: Fast and Effective Personalized Recommendation with Language Models
AI 摘要
论文提出 DuoLM 框架，将紧凑推荐模型与高效服务运行时结合：编码器 LLM 离线学习可复用的物品向量表示，个性化排序器 LLM 结合用户历史与反馈在线打分，并通过教师蒸馏端到端训练。运行时采用虚拟 token ID、共享历史计算与 CUDA Graph 重放等优化，在单张 H100 上每秒约 26,400 个候选打分，吞吐约等于使用 vLLM 与自动前缀缓存的教师模型，同时推荐质量与教师模型相当。
正文节选
DuoLM: Fast and Effective Personalized Recommendation with Language Models Abstract Language models can make personalized recommendations from rich item content, user history, and feedback. However, evaluating many candidates with full-text inputs incurs substantial online computation and limits serving throughput. We introduce DuoLM, a framework for fast and effective personalized recommendation that integrates a compact recommendation model with an efficient serving runtime. An encoder LLM lea