Xingdi (Eric) Yuan

Xingdi Yuan · 袁 星柢

Principal Researcher · Microsoft Research, Montréal

I aim to develop machines that can read, write, and use language as a tool, in both static datasets and interactive worlds.

At the Froggy team Froggy, my recent work focuses on teaching agentic coding systems to understand codebases, leverage memory, and to solve long-horizon tasks with tools. To achieve that, I create synthetic data, train models, build harnesses, and design evaluation pipelines.

Selected Publications

  • debug-gym: A Text-Based Environment for Interactive Debugging
    Xingdi Yuan*, Morgane M Moss, Charbel El Feghali, Chinmay Singh, Darya Moldavskaya, Drew MacPhee, Lucas Caccia, Matheus Pereira, Minseon Kim, Alessandro Sordoni, Marc-Alexandre Côté*
    2025
    Equipping an LLM-based coding agent with debuggers such as pdb.
    ArXiv
  • BugPilot: Complex Bug Generation for Efficient Learning of SWE Skills
    Atharv Sonwane*, Isadora White*, Hyunji Lee, Matheus Pereira, Lucas Caccia, Minseon Kim, Zhengyan Shi, Chinmay Singh, Alessandro Sordoni, Marc-Alexandre Côté, Xingdi Yuan
    2025
    Generating unintentional synthetic bugs by asking LLMs to design new features.
    ArXiv
  • Gistify! Codebase-Level Understanding via Runtime Execution
    Hyunji Lee, Minseon Kim, Chinmay Singh, Matheus Pereira, Atharv Sonwane, Isadora White, Elias Stengel-Eskin, Mohit Bansal, Zhengyan Shi, Alessandro Sordoni, Marc-Alexandre Côté, Xingdi Yuan, Lucas Caccia
    2025
    Generating an executable gist from a codebase and an entrypoint.
    ArXiv
  • Evolving Programmatic Skill Networks
    Haochen Shi, Xingdi Yuan*, Bang Liu*
    2025
    Forming a compositional network of skills (executable programs) that evolves through experience.
    ArXiv
  • Llama See, Llama Do: A Mechanistic Perspective on Contextual Entrainment and Distraction in LLMs Outstanding Paper
    Jingcheng Niu, Xingdi Yuan, Tong Wang, Hamidreza Saghir, Amir H. Abdi
    ACL 2025
    Using contextual entrainment to understand how LLMs become distracted by “irrelevant” contextual information in the prompt.
    ArXiv
  • Can Language Models Serve as Text-Based World Simulators?
    Ruoyao Wang, Graham Todd, Ziang Xiao, Xingdi Yuan, Marc-Alexandre Côté, Peter Clark, Peter Jansen
    ACL 2024
    Can LLMs predict p(s_t+1 | s_t, a_t) for text-based games?
    ArXiv
  • ALFWorld: Aligning Text and Embodied Environments for Interactive Learning
    Mohit Shridhar, Xingdi Yuan, Marc-Alexandre Côté, Yonatan Bisk, Adam Trischler, Matthew Hausknecht
    ICLR 2021
    Bridging ALFRED (visual) and TextWorld (text).
    ArXiv
  • Learning Dynamic Knowledge Graphs to Generalize on Text-Based Games
    Ashutosh Adhikari*, Xingdi Yuan*, Marc-Alexandre Côté*, Mikuláš Zelinka, Marc-Antoine Rondeau, Romain Laroche, Pascal Poupart, Jian Tang, Adam Trischler, William L. Hamilton
    NeurIPS 2020
    Building and maintaining a knowledge graph serves as an agent's belief.
    ArXiv
  • Interactive Fiction Games: A Colossal Adventure
    Matthew Hausknecht, Prithviraj Ammanabrolu, Marc-Alexandre Côté, Xingdi Yuan
    AAAI 2020
    Super hard text-based adventure games designed for human players.
    ArXiv
  • Interactive Machine Comprehension with Information Seeking Agents
    Xingdi Yuan*, Jie Fu*, Marc-Alexandre Côté, Yi Tay, Christopher Pal, Adam Trischler
    ACL 2020
    Gamifying any machine reading comprehension tasks by making the document partially observed, and equipping the agent with Ctrl+F.
    ArXiv
  • Interactive Language Learning by Question Answering
    Xingdi Yuan*, Marc-Alexandre Côté*, Jie Fu, Zhouhan Lin, Christopher Pal, Yoshua Bengio, Adam Trischler
    EMNLP 2019
    Training information seeking behaviors in text-based environments.
    ArXiv
  • TextWorld: A Learning Environment for Text-based Games
    Marc-Alexandre Côté, Ákos Kádár, Xingdi Yuan, Ben Kybartas, Tavian Barnes, Emery Fine, James Moore, Ruo Yu Tao, Matthew Hausknecht, Layla El Asri, Mahmoud Adada, Wendy Tay, Adam Trischler
    Computer Games Workshop, ICML/IJCAI 2018
    A sandbox learning environment for the training and evaluation of agents on text-based games.
    ArXiv

Experience

  • 2017 – Present
    Microsoft Research, Montréal
    Principal Researcher
  • 2015 – 2017
    Maluuba
    (Acquired by Microsoft)

Education

  • 2015
    M.S. Computer Science
    New York University
  • 2011
    B.S. Communications Engineering
    Beijing University of Technology
    北京工业大学

Service

  • Area Chair: ICML, NeurIPS, ARR
  • Action Editor: TACL
  • Outstanding Reviewer: EMNLP 2020, NeurIPS 2021, ICLR 2022, ICML 2022
  • Workshop Organizer: Wordplay Workshop (NeurIPS 2020, NAACL 2022, ACL 2024, EMNLP 2025), KBRL Workshop (IJCAI 2020)

Fun Stuff

asciiko (アス子)

A deep ascii art generator. Check out the YouTube demo or the open-sourced code.