PeopleofSuperintelligence

Researcher

Di He

Also known as 贺笛

Di He (贺笛) is a machine learning researcher whose work spans ranking theory, machine translation, graph neural networks, language models, and AI for science. At Peking University, he earned bachelor’s, master’s, and PhD degrees under Liwei Wang and worked as a senior researcher in Microsoft Research Asia’s Machine Learning Group. His team’s Graphormer won two 2021 challenges, and his work on graph neural network expressiveness won an ICLR Outstanding Paper Award in 2023 and received an ICLR Outstanding Paper Honorable Mention in 2024. He also coauthored a neural network framework for variational Monte Carlo and a token-level method for reinforcement learning from human feedback. As of October 2026, he is an assistant professor at Peking University, where his research focuses on generative models and AI for science.

Works
6
Citations
3
h-index
1

Publication numbers from OpenAlex, as of 19 October 2026.

Timeline

  1. 2013

    NDCG ranking theory

    In 2013, He coauthors *A Theoretical Analysis of NDCG Type Ranking Measures*, which studies how NDCG ranking measures behave as the number of items grows.

  2. 2016

    Dual learning for translation

    In 2016, he coauthors *Dual Learning for Machine Translation*. Its dual-learning game trains neural machine translation systems from unlabeled data.

  3. 2021

    Graphormer wins two contests

    His team designs Graphormer, which wins the KDD 2021 molecular-property prediction challenge and the NeurIPS 2021 molecular-dynamics simulation challenge.

  4. 2023

    An ICLR award for biconnectivity

    He coauthors *Rethinking the Expressive Power of GNNs via Graph Biconnectivity*, which studies biconnectivity as an expressivity metric for graph neural networks. The paper wins an ICLR 2023 Outstanding Paper Award.

  5. 2024

    Quantifying GNN expressiveness

    The paper defines homomorphism expressivity as a measure of GNNs’ ability to count graphs under homomorphism. It receives an ICLR 2024 Outstanding Paper Honorable Mention.

  6. 2024

    Neural variational Monte Carlo

    He coauthors a neural network-based variational Monte Carlo framework named Forward Laplacian, which computes neural-network Laplacians through efficient forward propagation. *Nature Machine Intelligence* publishes the work in February 2024.

  7. 2025

    Token optimization for RLHF

    The paper appears at ICML 2025 with He among its authors. It introduces Reinforced Token Optimization, which learns token-level rewards from preference data and combines direct preference optimization with PPO.

  8. As of October 2026

    Assistant Professor at Peking University

    As of October 2026, he is an assistant professor in Peking University’s School of Intelligence Science and Technology, where he studies generative models and AI for science. He earns bachelor’s, master’s, and PhD degrees at Peking University under Liwei Wang and previously serves as a Senior Researcher in Microsoft Research Asia’s Machine Learning Group.

Career

  1. Since 2025Peking UniversityAssistant Professor
  2. Microsoft ResearchSenior Researcher

Education

  1. Peking UniversityBachelor’s degree

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

Most-cited works

  1. A Mathematical Formulation of AGI in the (C, U, V) Framework2025 · Engineering · 3 citationsWith Song-Chun Zhu
  2. AgentFactory: A Self-Evolving Framework Through Executable Subagent Accumulation and Reuse2026 · Annual Meeting of the Association for Computational Linguistics (ACL) · 0 citations
  3. Efficient Reasoning for Large Reasoning Language Models via Certainty-Guided Reflection Suppression2026 · Proceedings of the AAAI Conference on Artificial Intelligence · 0 citations
  4. Enhancing Auto-regressive Chain-of-Thought through Loop-Aligned Reasoning2026 · Conference of the European Chapter of the Association for Computational Linguistics · 0 citations
  5. SWE-Swiss: A Multi-Task Fine-Tuning and RL Recipe for High-Performance Issue Resolution2026 · 0 citations

From OpenAlex, as of 19 October 2026.

  1. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionIncludes Peking University. Our estimates of where the people shaping AI studied and where their careers led, from education records in cited biographies.
  2. The company share of AI researchers’ top papers rose from 2% to 32%Includes Peking University. We estimate where leading AI researchers worked when their most-cited papers appeared. The company share rose from 2% in 1975–1979 to 32% in 2020–2024.
  3. Google has the highest combined h-index among AI organizationsIncludes Peking University. Our estimates of combined and median h-index at AI organizations, using current affiliations and OpenAlex records.
  4. Stanford University trained or employed 20 of the 89 AI major-prize winners, more than any other universityIncludes Peking University. Stanford University trained or employed 20 of the 89 AI major-prize winners, more than any other university. We compare doctorate and employment links for major-prize and best-paper winners.

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 Di He yet. Send us a free-licensed photo.

  1. A Theoretical Analysis of NDCG Type Ranking Measures proceedings.mlr.press. Checked 19 October 2026.
  2. [1611.00179] Dual Learning for Machine Translation arxiv.org. Checked 19 October 2026.
  3. Di He-北京大学智能学院英文 sai.pku.edu.cn. Checked 19 October 2026.
  4. Announcing the ICLR 2023 Outstanding Paper Award Recipients – ICLR Blog blog.iclr.cc. Checked 19 October 2026.
  5. ICLR 2024 Outstanding Paper Awards – ICLR Blog blog.iclr.cc. Checked 19 October 2026.
  6. [2401.08514] Beyond Weisfeiler-Lehman: A Quantitative Framework for GNN Expressiveness arxiv.org. Checked 19 October 2026.
  7. A computational framework for neural network-based variational Monte Carlo with Forward Laplacian | Nature Machine Intelligence nature.com. Checked 19 October 2026.
  8. DPO Meets PPO: Reinforced Token Optimization for RLHF - Microsoft Research microsoft.com. Checked 19 October 2026.
  9. Di He-北京大学人工智能研究院 ai.pku.edu.cn. Checked 19 October 2026.
  10. About me - Di He (贺笛) dihe-pku.github.io. Checked 19 October 2026.