Shangjian Yin
Hi! My name is Shangjian Yin. I'm a CS PhD student at the University of California, Riverside, advised by Zhouxing Shi. Previously, I was a Research Intern at Microsoft AI and Meta AI.
Currently, I focus on LLM post-training, recursive self-improving LLMs, and long-horizon agents. I am also interested in world models, Physical AI, and agent systems infra, and I am open to collaboration!
Selected First-Author Research
Recursive Self-Improvement via On-Policy Distillation for Reasoning
GRLO: Towards Generalizable Reinforcement Learning in Open-Ended Environments from Zero
Rethinking Reasoning Post-Training with General Chat Boosting and Dual-Reward Refinement
From Individual to Common: An Early Exploration of Consensus in Non-verifiable Data for Preference Optimization
Align Large Language Model with Human Preference via Extremely Self-Synthetic Data
PIKA: Expert-Level Synthetic Datasets for Post-Training Alignment from Scratch
ECLM: Entity-Level Large Language Model for Spoken Language Understanding with Chain of Intent
MIDLM: Multi-Intent Detection with Bidirectional Large Language Models
Uni-MIS: United Multiple Intent Spoken Language Understanding via Multi-View Intent-Slot Interaction
Latest update: 09/2026.