PeopleofSuperintelligence

Researcher

Yadong Lu

Also known as Adam Lu

Yadong Lu is a principal researcher at Microsoft Research whose work spans synthetic training data and agents that operate through visual interfaces, Windows, and the web. In 2018, while pursuing a statistics PhD at UC Irvine, he coauthored an award-winning code-clone detection paper; in 2024, he coauthored AgentInstruct, Windows Agent Arena, and OmniParser. His later projects scaled multimodal web-agent trajectories, gave web agents a terminal to explore sites through code, and built FaraGen1.5’s data pipeline from environments, solvers, and verifiers. As of October 2026, he works in Redmond on computer-use agents and efficient training methods for large vision-language models.

Works
17
Citations
67
h-index
4

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

  1. 2018

    Oreo wins a paper award

    In 2018, while a statistics PhD student at UC Irvine advised by Pierre Baldi, he coauthors “Oreo: Detection of Clones in the Twilight Zone,” which receives the ESEC/FSE Distinguished Paper Award. The detector combines machine learning, information retrieval, and software metrics to find difficult code clones.

  2. July 2024

    AgentInstruct for post-training

    In July 2024, he coauthors AgentInstruct, an agentic framework for synthetic language-model post-training. It creates 25 million prompt-response pairs for teaching language models through instruction tuning.

  3. September 2024

    Windows Agent Arena

    In September 2024, he coauthors Windows Agent Arena, a reproducible, open-source Windows benchmark with 150+ tasks that test agent planning, screen understanding, and tool use.

  4. October 2024

    OmniParser for visual agents

    In October 2024, he coauthors OmniParser, a compact screen parser that turns UI screenshots into structured elements. Paired with GPT-4V, it improves how agents ground actions to interface regions.

  5. February 2025

    Explorer web trajectories

    In February 2025, he coauthors Explorer, a method for synthesizing multimodal web-agent data. Its dataset includes over 94,000 successful trajectories across 49,000 unique URLs.

  6. May 2026

    Webwright terminal harness

    In May 2026, he coauthors Webwright, a web-agent harness that gives a model a terminal to spawn browser sessions and explore websites through code.

  7. June 2026

    FaraGen1.5 data pipeline

    In June 2026, he coauthors Fara-1.5, which introduces FaraGen1.5, a scalable data pipeline for computer-use agents built from environments, solvers, and verifiers.

  8. As of October 2026

    Principal researcher at Microsoft

    As of October 2026, he is a principal researcher at Microsoft Research in Redmond, working on computer-use agents and efficient pre/post-training methods for large vision-language models.

Career

  1. NowMicrosoft ResearchPrincipal researcher
  2. 2008–2010China Aerospace Science and Technology Corporation

Education

  1. University of California, IrvinePhD student in statistics

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

  1. Among employers with 30+ full-time cases, OpenAI leads primary technical role H-1B pay at $310,000, and Amazon, Google and Microsoft file 57% of all casesIncludes Microsoft. Across 440,669 certified US H-1B filings from 44 employers (FY2020–FY2026), OpenAI leads primary technical role base pay at $310,000 and xAI leads overall base pay at $290,000.
  2. Google had the most people shaping AI every year from 2014 to 2025Includes Microsoft. Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  3. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Microsoft. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.
  4. Google has the highest combined h-index among AI organizationsIncludes Microsoft. We estimate that Google holds the highest combined h-index at 8,016, while universities account for 73% of the deduplicated total across AI organizations.

Sources and credits

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

  1. Multidepartmental Collaboration on Detecting Code Clones Leads to Distinguished Paper Award ics.uci.edu. Checked 22 October 2026.
  2. AgentInstruct: Toward Generative Teaching with Agentic Flows arxiv.org. Checked 22 October 2026.
  3. Windows Agent Arena: Evaluating Multi-Modal OS Agents at Scale arxiv.org. Checked 22 October 2026.
  4. Windows Agent Arena microsoft.github.io. Checked 22 October 2026.
  5. OmniParser for pure vision-based GUI agent microsoft.com. Checked 22 October 2026.
  6. Explorer: Scaling Exploration-driven Web Trajectory Synthesis for Multimodal Web Agents arxiv.org. Checked 22 October 2026.
  7. Webwright: A Terminal Is All You Need For Web Agents microsoft.com. Checked 22 October 2026.
  8. Fara-1.5: Scalable Learning Environments for Computer Use Agents arxiv.org. Checked 22 October 2026.
  9. Yadong Lu at Microsoft Research microsoft.com. Checked 22 October 2026.
  10. Yadong (Adam) Lu adamlu123.github.io. Checked 22 October 2026.