PeopleofSuperintelligence

Researcher

Lingxiao Zhao

Lingxiao Zhao is an AI scientist at Mistral AI who works on generative models and the pretraining of large language models. Their path into AI began with an electrical engineering degree from Xi'an Jiaotong University and led to a joint PhD in machine learning and public policy at Carnegie Mellon under Leman Akoglu, finished with a 2024 dissertation on graph representation learning. During the doctorate, they wrote PairNorm, a fix for oversmoothing in deep graph neural networks, and the GNN-AK framework for more expressive GNNs. At Zoom from 2024 to 2025, they co-authored Chain of Draft, a prompting method that lets language models reason with far fewer tokens. Zhao joined Mistral in April 2025 and co-authored the reports on Magistral, the company's first reasoning model, and on the Ministral 3 and Shieldstral models.

Works
29
Citations
1,146
h-index
11
Badges

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

  1. 2012–2016

    From Xi'an Jiaotong to CMU

    Zhao studies electrical engineering at Xi'an Jiaotong University from 2012 to 2016, working in the power systems area. In 2016, they begin a master's degree in the ECE department at Carnegie Mellon University.

  2. 2018–2020

    PairNorm for deep GNNs

    In 2018, they begin a PhD in machine learning and public policy, run jointly by Carnegie Mellon's Machine Learning Department and Heinz College, with Leman Akoglu as advisor. With Akoglu, they write PairNorm, a normalization layer that keeps node embeddings in deep graph neural networks from becoming too similar; it is accepted at ICLR 2020.

  3. 2021–2022

    The GNN-AK framework

    In October 2021, with Wei Jin, Leman Akoglu and Neil Shah, they post From Stars to Subgraphs, which introduces GNN-AK, a framework that makes any message-passing GNN more expressive by extending local aggregation from star patterns to subgraph patterns such as k-egonets. The paper and Graph Condensation for Graph Neural Networks are accepted at ICLR 2022.

  4. 2021–2023

    Internships at Snap, NVIDIA and Apple

    During the PhD, they work as a research intern at Snap Inc. in 2021, at NVIDIA in 2022 and at Apple in 2023.

  5. 2024

    Pard and a CMU dissertation

    Pard, posted in February 2024 with Xueying Ding and Leman Akoglu, is a permutation-invariant autoregressive diffusion model for graph generation. In 2024, they complete the dissertation Towards More Powerful Graph Representation Learning at Carnegie Mellon.

  6. April 2024–April 2025

    Research scientist at Zoom

    Zhao works as a research scientist at Zoom from April 2024 to April 2025. In December 2024, they co-author T-REG, a preference-optimization method for language models that combines sequence-level and token-level rewards. Chain of Draft, which they co-author in February 2025, has models write terse intermediate reasoning and matches or beats chain-of-thought accuracy with as little as 7.6% of the tokens.

  7. 2025

    Magistral at Mistral AI

    They join Mistral AI as an AI scientist in April 2025. In June 2025, they are among the co-authors of Magistral, Mistral's first reasoning model, trained with the company's own reinforcement learning pipeline; Magistral Small is open-sourced under the Apache 2.0 license.

  8. As of October 2026

    AI scientist at Mistral AI

    In January 2026, they co-author the report on Ministral 3, a family of 3B, 8B and 14B dense language models with image understanding. Shieldstral, a 3B-parameter multimodal safety classifier that treats content moderation as a yes-or-no question-answering task, follows in July 2026 with Zhao among its co-authors. They are an AI scientist at Mistral AI, focused on generative models and LLM pretraining.

Career

  1. Since 2025Mistral AIAI scientist
  2. 2024–2025ZoomResearch scientist
  3. 2023AppleResearch intern
  4. 2023–2025Dalian Maritime University
  5. 2022NVIDIAResearch intern
  6. 2021SnapResearch intern

Education

  1. 2024Carnegie Mellon UniversityPhD
  2. 2016Xi'an Jiaotong UniversityElectrical engineering degree
  3. From 2016Carnegie Mellon UniversityMaster'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. Beyond Homophily in Graph Neural Networks: Current Limitations and Effective Designs2020 · 492 citations
  2. PairNorm: Tackling Oversmoothing in GNNs2019 · 278 citations
  3. Graph Unrolling Networks: Interpretable Neural Networks for Graph Signal Denoising2021 · IEEE Transactions on Signal Processing · 83 citations
  4. On Using Classification Datasets to Evaluate Graph Outlier Detection: Peculiar Observations and New Insights2021 · Big Data · 58 citations
  5. Supervised locally linear embedding with probability-based distance for classification2008 · Computers & Mathematics with Applications · 48 citations

From OpenAlex, as of 21 October 2026.

  1. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Lingxiao Zhao. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.
  2. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Lingxiao Zhao. We estimate that alumni of 17 major AI labs went on to found or lead 350 other organizations, including 226 they founded, and we trace 472 career moves directly between 23 labs since 2010.
  3. Companies supply the most prior-work hires at 7 of the 9 frontier AI labsIncludes Mistral AI. Our estimates of the organizations where 550 people worked directly before joining 9 frontier AI labs across 620 career transitions, comparing the main feeder organizations per lab.
  4. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Mistral AI. Our estimates of where the 943 people shaping AI who worked at Google and Google DeepMind work now: 446 at Google and 452 across 278 other organizations.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

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

  1. Lingxiao Zhao lingxiaozhao.com. Checked 20 October 2026.
  2. Lingxiao Zhao (Mistral AI): pretraining LLMs — Lacuna lacuna.tiptreesystems.com. Checked 20 October 2026.
  3. Lingxiao Zhao | alphaXiv alphaxiv.org. Checked 20 October 2026.
  4. PairNorm: Tackling Oversmoothing in GNNs arxiv.org. Checked 20 October 2026.
  5. From Stars to Subgraphs: Uplifting Any GNN with Local Structure Awareness arxiv.org. Checked 20 October 2026.
  6. Pard: Permutation-Invariant Autoregressive Diffusion for Graph Generation arxiv.org. Checked 20 October 2026.
  7. Ph.D. Dissertations - Machine Learning Department, Carnegie Mellon University ml.cmu.edu. Checked 20 October 2026.
  8. T-REG: Preference Optimization with Token-Level Reward Regularization arxiv.org. Checked 20 October 2026.
  9. Chain of Draft: Thinking Faster by Writing Less arxiv.org. Checked 20 October 2026.
  10. Magistral arxiv.org. Checked 20 October 2026.
  11. Ministral 3 arxiv.org. Checked 20 October 2026.
  12. Shieldstral arxiv.org. Checked 20 October 2026.