Chong Wang

Chong Wang is a machine-learning researcher who has led applied AI research and developed methods for speech and language models. He grew up in Liaoning, earned bachelor’s and master’s degrees at Tsinghua University in 2006, completed a Princeton computer-science PhD in 2012, and worked in Carnegie Mellon’s Machine Learning Department. With David Blei, he published a topic model for scientific-article recommendations in 2011; his research career includes roles at Google and Microsoft Research and leadership of ByteDance’s applied-machine-learning team, whose work supports TikTok and Douyin. SIGKDD awarded the paper its Test of Time Award in 2021; his Apple work includes foundation models, RATTENTION, and instruction-following pruning, and he joined Meta in 2025, where he is a research scientist.

Now at: Meta
Works: 118

Publication numbers from OpenAlex, as of 19 October 2026. Citations and h-index are hidden because this OpenAlex record may be incomplete or mix in other authors.

Timeline

2006

Tsinghua degrees

He completes his bachelor's and master's degrees at Tsinghua University in 2006 after growing up in Liaoning, China.
2011

Collaborative topic modeling

With David Blei, he publishes a collaborative topic model for recommending scientific articles at KDD in 2011.
2012–2013

A Princeton PhD and CMU role

He earns a computer-science PhD at Princeton in 2012 and works as a project scientist in Carnegie Mellon's Machine Learning Department in 2013, studying probabilistic graphical models.
2018

UIS-RNN speaker diarization

He co-authors UIS-RNN in 2018, a fully supervised speaker-diarization system that uses a parameter-sharing recurrent neural network for each speaker.
By 2021

Applied machine learning at ByteDance

Before ByteDance, he works as a research scientist at Google and Microsoft Research. By 2021, he leads ByteDance's applied machine-learning research team, whose work supports products including TikTok and Douyin. In 2021, his 2011 paper with David Blei receives SIGKDD's Test of Time Award.
2024

Apple Intelligence foundation models

At Apple, he co-authors Apple Intelligence Foundation Language Models in 2024, describing on-device and server models that power Apple Intelligence.
2025

Efficient language model attention

In 2025, he co-authors RATTENTION, which adds linear attention to local attention to capture information outside the window. He also co-authors instruction-following pruning, which uses user instructions to select model parameters.
Since 2025

Research scientist at Meta

He leaves Apple for Meta in 2025; he is a research scientist there.