Yi Zhang
Yi Zhang completed a computer science Ph.D. at Princeton University in 2021, where he received the Wallace Memorial Fellowship in Engineering, after bachelor's degrees from Shanghai Jiao Tong University and the University of Michigan, Ann Arbor. His research concentrated on theoretical machine learning. In 2023, as a senior researcher at Microsoft Research, he coauthored Textbooks Are All You Need, which introduced phi-1, a Transformer-based model with 1.3B parameters, and Sparks of Artificial General Intelligence, an investigation of an early version of GPT-4. In 2024 he coauthored the Phi-4 technical report on a 14-billion parameter language model developed with a training recipe that is centrally focused on data quality. In April 2025 he coauthored CASCADE with Runlong Zhou, with an affiliation at Apple.
Timeline
-
2017
Invertibility of convolutional nets
He coauthors Towards Understanding Invertibility of Convolutional Neural Networks with Anna Gilbert, Kibok Lee, Yuting Zhang, and Honglak Lee at IJCAI in 2017.
-
2018
Generalization bounds for deep nets
In 2018, Sanjeev Arora, Rong Ge, Behnam Neyshabur, and he publish work on stronger generalization for deep nets via a compression approach.
-
2021
A Princeton computer science doctorate
He obtains a Ph.D. in computer science from Princeton University in 2021, where he receives the Wallace Memorial Fellowship in Engineering. Prior to Princeton, he receives one bachelor's degree from Shanghai Jiao Tong University and another from the University of Michigan, Ann Arbor. His research concentrates on theoretical machine learning.
-
March 2023
Early experiments with GPT-4
He coauthors Sparks of Artificial General Intelligence: Early experiments with GPT-4, submitted in 2023. The authors investigate an early version of GPT-4.
-
April 2023
Senior researcher at Microsoft
In April 2023 he is a senior researcher at Microsoft Research and discusses his recent work on Sparks of Artificial General Intelligence: Early experiments with GPT-4.
-
June 2023
Textbooks are all you need
He coauthors Textbooks Are All You Need with Suriya Gunasekar and Jyoti Aneja in 2023. The paper introduces phi-1, a Transformer-based model with 1.3B parameters, trained on textbook quality data and synthetically generated textbooks, which attains pass@1 accuracy 50.6% on HumanEval and 55.5% on MBPP.
-
December 2024
The Phi-4 language model
In 2024 he coauthors the Phi-4 technical report. It presents phi-4, a 14-billion parameter language model developed with a training recipe that is centrally focused on data quality.
-
April 2025
Research affiliation at Apple
In April 2025 he coauthors CASCADE Your Datasets for Cross-Mode Knowledge Retrieval of Language Models with Runlong Zhou, and his affiliation is Apple. The paper proposes CASCADE, a pretraining algorithm that uses cascading datasets with varying sequence lengths.
Sources and credits
Photo: We have no published photo of Yi Zhang yet. Send us a free-licensed photo.
- Yi Zhang - Academic Homepage
- Stronger Generalization Bounds for Deep Nets via a Compression Approach
- Guest Speakers: Machine Learning - Elmore Family School of Electrical and Computer Engineering - Purdue University
- [2303.12712] Sparks of Artificial General Intelligence: Early experiments with GPT-4
- [2306.11644] Textbooks Are All You Need
- Textbooks Are All You Need - Microsoft Research
- [2412.08905] Phi-4 Technical Report
- Phi-4 Technical Report - Microsoft Research
- CASCADE Your Datasets for Cross-Mode Knowledge Retrieval of Language Models