Yue Wu

Yue Wu is a machine-learning researcher at Carnegie Mellon University whose work connects language models with agents that learn from written guides and interactions. He earned a computer science degree at CMU, then worked on a graph-based recommender system at IBM Research and game agents at Microsoft Research. In 2024, he coauthored AgentKit and Microsoft’s Phi-4 report; in 2025, he coauthored Phi-4-reasoning. As of October 2026, he is a PhD student in CMU’s Machine Learning Department, working on foundational models and agents.

Now at: Carnegie Mellon University
Badges: Bridge

Timeline

2017–2020

A CMU computer science degree

He earns a B.S. in computer science and a minor in machine learning at Carnegie Mellon University between 2017 and 2020.
2022

An IBM research internship

At IBM Research, he works as a research intern in summer 2022 on a graph-based recommender system for long, unstructured text using large language models.
February 2023

Reading Atari manuals

In February 2023, he coauthors Read and Reap the Rewards, a framework that helps Atari reinforcement-learning agents use developer manuals and Wikipedia articles. A question-answering module extracts manual information, and an LLM reasons about agent-object interactions to assign auxiliary rewards.
2023

Microsoft Research game agents

He interns at Microsoft Research in summer 2023 on SmartPlay. In fall 2023, he consults on GPT and LLaMA agents for games.
2024

AgentKit for LLM agents

He coauthors AgentKit in 2024, a prompting framework for multifunctional LLM agents that connects task-specific prompts into reasoning chains.
December 2024

Coauthoring the Phi-4 report

He coauthors Microsoft Research’s Phi-4 Technical Report in December 2024; it introduces a 14-billion-parameter language model with a training recipe centered on data quality and synthetic data.
April 2025

Phi-4-reasoning for complex tasks

He coauthors Phi-4-reasoning in April 2025, a 14-billion-parameter model fine-tuned on curated prompts and reasoning demonstrations generated with o3-mini. The report also introduces Phi-4-reasoning-plus, which adds outcome-based reinforcement learning.
As of October 2026

PhD student at Carnegie Mellon

As of October 2026, he is a PhD student in Carnegie Mellon’s Machine Learning Department, working on foundational models, agents and machine learning.