Welcome back to the fifth installment of A Random Walk Down Recsys. This batch of papers reflects how quickly the generative recommendation playbook is being adapted to new verticals and tightened ...

Welcome back to the fourth installment of A Random Walk Down Recsys. This time, the three papers span a range of practical challenges in generative recommendation: efficiently compressing long user...

Welcome back to the third installment of A Random Walk Down Recsys. This time, all five papers revolve around a single theme: Semantic IDs (SIDs) — how to generate them, how to improve their qualit...

Welcome back to the second installment of A Random Walk Down Recsys. In this post, I continue surveying interesting papers from the Arxiv IR section, covering five recent works: HyFormer, Token-lev...

This post documents my journey learning VERL (Volcano Engine Reinforcement Learning), a scalable and efficient reinforcement learning framework, through the lens of OpenOneRec’s implementation. Ope...

This is a new series of blog beyond my conference paper reading blog, in which I would summarize the paper that I found interesting form Arxiv IR section and share my learnings. In this first blog...

Fully Sharded Data Parallel (FSDP) is PyTorch’s approach to training large models that don’t fit in a single GPU’s memory. FSDP2 represents a significant redesign from FSDP1, with improved performa...

In this post, I would like to share some insights from the paper I have read in year 2025 and summarize some trends over the year. The One The best work I enjoyed this year is the One-series from...

In this post, I would like to summary the paper from Recsys 2025 and share some of my learnings. We would cover several topics such as sequence modeling, cross domain learning as well as LLM integr...

It has been a while since KDD 2025, finally I have had sometime to finish reading all papers that I interested in and summarize some of my learnings in this post . My primary focus is ...