PeopleofSuperintelligence

Researcher

Dingli Yu

Dingli Yu is an AI researcher whose work spans language-model theory, evaluation, and development. In Tsinghua's Yao Class, he studied computer science, won gold medals at the IOI and ACM-ICPC World Finals, and completed a Princeton PhD under Sanjeev Arora. His research includes neural-tangent-kernel analysis of language-model fine-tuning, the Skill-Mix evaluation, work on skill composition, and Microsoft's Phi-4. He joined OpenAI in February 2025 and is a Member of Technical Staff focused on large language models and synthetic data generation.

Now at
OpenAI
Works
206
Citations
2,789
h-index
29
Badges

Publication numbers from OpenAlex, as of 19 October 2026.

Timeline

  1. 2014

    An IOI gold medal

    At the 2014 IOI, he wins a gold medal and places seventh in the individual results.

  2. 2014–2018

    Tsinghua's Yao Class

    At Tsinghua's Institute for Interdisciplinary Information Science, he studies computer science in Yao Class and graduates with the 2018 cohort.

  3. 2015

    An ICPC World Finals gold

    He wins an ACM-ICPC World Finals gold medal in 2015 with Tsinghua's ThinkingBear team, which finishes fourth.

  4. 2023

    Kernel theory and Skill-Mix

    In 2023, he coauthors A Kernel-Based View of Language Model Fine-Tuning, a study of when neural tangent kernels can describe fine-tuning of pre-trained language models. He also coauthors Skill-Mix, an evaluation that asks models to create text using randomly selected combinations of skills.

  5. 2024

    A Princeton PhD

    At Princeton, he completes a computer science PhD under Sanjeev Arora in 2024. In July 2024, he presents his final public oral, “Efficient Scaling of Large Models: Principles in Optimization and Data Aspects,” on neural-network training dynamics and large-model scaling.

  6. 2024

    Learning skill combinations

    At NeurIPS 2024, he coauthors a study of whether examples can teach smaller models to combine skills. Fine-tuning 7B and 13B models on combinations of two and three skills improves their ability to compose four and five skills.

  7. 2024

    Phi-4 at Microsoft

    He helps develop Phi-4 at Microsoft and coauthors its December 2024 technical report. The report presents a 14-billion-parameter language model that uses synthetic data throughout training.

  8. Since 2025

    Research at OpenAI

    In February 2025, he joins OpenAI and researches large language models, with a focus on synthetic data generation. He is a Member of Technical Staff.

Career

  1. Since 2025OpenAIMember of Technical Staff
  2. 2024–2025Northwestern Polytechnical University
  3. 2019–2023Jilin University
  4. 1996–2025Liverpool John Moores University

Education

  1. 2024Princeton UniversityPhD in Computer Science
  2. 2018Tsinghua University

From the biography and OpenAlex affiliations (the years of papers that list the organization), so some years are approximate.

Most-cited works

  1. Selecting radial basis function network centers with recursive orthogonal least squares training2000 · IEEE Transactions on Neural Networks · 242 citations
  2. Adaptive neural network model based predictive control for air–fuel ratio of SI engines2005 · Engineering Applications of Artificial Intelligence · 139 citations
  3. Implementation of neural network predictive control to a multivariable chemical reactor2003 · Control Engineering Practice · 104 citations
  4. Harnessing the Power of Infinitely Wide Deep Nets on Small-data Tasks2019 · 91 citationsWith Ruslan Salakhutdinov and Sanjeev Arora
  5. Automatic visual defects inspection of wind turbine blades via YOLO-based small object detection approach2019 · Journal of Electronic Imaging · 84 citations

From OpenAlex, as of 19 October 2026.

  1. The median h-index is 53 at universities and 19 at frontier AI labsIncludes Dingli Yu. The median h-index is 53 at universities and 19 at frontier AI labs. We compare publishing careers, citations and moves to estimate how universities and frontier labs contribute to AI research.
  2. 14% of AI career moves in the 2020s went from a university to a companyIncludes Dingli Yu. Our estimates of how university-to-company career transitions changed by decade, including roles added alongside an earlier job.
  3. 13 of 79 people shaping AI with a former OpenAI role now work at AnthropicIncludes OpenAI. Our estimates of where OpenAI recruits came from and where people with former OpenAI roles work now.
  4. Google alumni founded or led 87 other organizations, the most among the labs we compareIncludes OpenAI. We trace 212 organizations founded or led by AI lab alumni, and 223 career moves between the labs.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Dingli Yu yet. Send us a free-licensed photo.

  1. Dingli Yu dingliyu.net. Checked 19 October 2026.
  2. Profile of Dingli Yu - Competitive Programming Hall Of Fame cphof.org. Checked 19 October 2026.
  3. IIIS holds Commencement and YaoClass Meeting 2018 iiis.tsinghua.edu.cn. Checked 19 October 2026.
  4. A Kernel-Based View of Language Model Fine-Tuning proceedings.mlr.press. Checked 19 October 2026.
  5. Skill-Mix: a Flexible and Expandable Family of Evaluations for AI models arxiv.org. Checked 19 October 2026.
  6. 07-18 Dingli Yu FPO | CS cs.princeton.edu. Checked 19 October 2026.
  7. Current Members | Arora Research Lab aroraresearch-princeton.github.io. Checked 19 October 2026.
  8. Can Models Learn Skill Composition from Examples? | NeurIPS Proceedings proceedings.neurips.cc. Checked 19 October 2026.
  9. Phi-4 Technical Report - Microsoft Research microsoft.com. Checked 19 October 2026.