AI People Wiki

Builders & open source

John Yang

Computer Science PhD student, Stanford University

John Yang is a computer science PhD student at Stanford University, advised by Ludwig Schmidt and Diyi Yang, researching language model agents, evaluation, synthetic data and software engineering. Yang's site describes Yang as the creator of InterCode, SWE-bench, SWE-agent, SWE-smith and CodeClash, and also features ProgramBench. Yang is the last-listed author of SWE-sweep, a bug-finding and repair benchmark released by Meta, whose project page lists Meta Superintelligence Labs among Yang's affiliations.123

At a glance

Current role
Computer Science PhD student
Sector
University
Education
Princeton, Stanford
Work listed
3
Sources cited
6
Profile reviewed

Research

Yang completed an M.S.E. in computer science at Princeton University, advised by Karthik Narasimhan, and is now a PhD student at Stanford University advised by Ludwig Schmidt and Diyi Yang. Yang's research covers language model agents, language model evaluation, synthetic data and software engineering.1

Yang co-authored SWE-bench (October 2023), which tests whether language models can resolve real GitHub issues, and was first author of SWE-agent (May 2024) and SWE-smith (April 2025). Yang's site also lists InterCode and CodeClash among the projects Yang created and features ProgramBench.4561

SWE-sweep

Yang is the final author of SWE-sweep, released with Kilian Lieret as first author. The benchmark gives agents a real repository and asks them to find and fix as many bugs as possible, scoring hidden per-bug tests and assigning zero to a task if previously passing tests break.23

Work and contributions

  • SWE-sweep Open-source code

    Last-listed author of SWE-sweep, a benchmark in which agents must discover and repair bugs in real repositories without hints about bug type or location.23

  • SWE-bench Open data

    Co-author of "SWE-bench: Can Language Models Resolve Real-World GitHub Issues?" (2023); Yang's site lists Yang as the creator of SWE-bench and its Verified, Multimodal and Multilingual variants.41

  • SWE-agent and SWE-smith Open-source code

    First author of SWE-agent (2024), on agent-computer interfaces for software engineering, and SWE-smith (2025), on scaling training data for software engineering agents.56

Timeline

  1. Co-authored SWE-bench, a benchmark of real-world GitHub issues for language models.4
  2. Posted SWE-agent, on agent-computer interfaces for software engineering, as first author.5
  3. Posted SWE-smith, on scaling data for software engineering agents, as first author.6
  4. Co-authored SWE-sweep, a bug discovery and repair benchmark released by Meta, as last author.3

Other documented roles

Roles the sources document that are not confirmed as current.

Where they work

Suggest a correction or removal

Profiles cover public professional work only, and each statement cites a source. If something is out of date, missing or wrong, tell us and link a source. People can also ask for their profile to be corrected or removed.

References

6 sources · profile reviewed Oct 22, 2026 · data updated Oct 22, 2026

  1. John Yang — personal siteJohn Yang · Official page · accessed Oct 22, 2026
  2. SWE-sweep — project site (authors and affiliations)SWE-sweep · Official page · accessed Oct 22, 2026
  3. facebookresearch/swe-sweep README (citation and evaluation)facebookresearch (GitHub repository) · Repository · accessed Oct 22, 2026
  4. SWE-bench: Can Language Models Resolve Real-World GitHub Issues? (arXiv 2310.06770)arXiv · Paper · published Oct 10, 2023 · accessed Oct 22, 2026
  5. SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering (arXiv 2405.15793)arXiv · Paper · published May 6, 2024 · accessed Oct 22, 2026
  6. SWE-smith: Scaling Data for Software Engineering Agents (arXiv 2504.21798)arXiv · Paper · published Apr 30, 2025 · accessed Oct 22, 2026