Xingyao Wang (王星尧)

Pronouns: he/him/his. How to pronounce?

I am a second-year PhD student in the Department of Computer Science at the University of Illinois Urbana-Champaign, advised by Prof. Heng Ji. I also work closely with Prof. Hao Peng.

I’m interested in developing interactive language agents powered by foundation models like large language models (LLMs) and vision-language models (VLMs). These agents can interact with both computer and physical environments through programming languages, assisting humans and continuously self-improving based on environmental observations and human language feedback.

Beyond research, I contribute to open-source software and am a core contributor to OpenDevin Stargazers .

I received my undergraduate degree at the University of Michigan in computer science and data science. Previously, I was an intern at Google Research (2023, multimodal LLM pre-training), Microsoft (2022, unit-test generation), Bytedance (2021, lightseq Stargazers ), and Tencent (2020, distributed deep learning).

news


2024 Jun Excited to share that R-Tuning receives outstanding paper award from NAACL 2024!
2024 Apr Check out our newly released paper, introducing a SOTA reasoning generalist Eurus model and UltraInteract (data)! UltraInteract collects trees (We love trees 🌲!) of multi-turn interactions with preference pairs that support both SFT and preference learning for challenging reasoning problems!
2024 Feb LLM agents act better when they take actions by generating Python code: Check out our new reprint on an LLM agent (CodeAct) deeply integrated with Python! Feel free to chat with our demo here!
2024 Jan MINT and CRAFT are accepted to ICLR 2024! See y’all in Vienna this May!
2023 May Excited to share that I’ll be joining Google Research NYC this summer as a student researcher!

selected publications

2024

  1. Executable Code Actions Elicit Better LLM Agents
    ICML 2024; LLMAgents Workshop @ ICLR 2024 (Oral), Feb 2024
  2. A Single Transformer for Scalable Vision-Language Modeling
    Yangyi Chen*Xingyao Wang*Hao Peng, and Heng Ji
    Arxiv preprint, Jul 2024
  3. MINT: Evaluating LLMs in Multi-turn Interaction with Tools and Language Feedback
    In Proceedings of the International Conference on Learning Representations, May 2024
  4. CRAFT: Customizing LLMs by Creating and Retrieving from Specialized Toolsets
    Lifan Yuan*Yangyi Chen*Xingyao WangYi R. FungHao Peng, and Heng Ji
    In Proceedings of the International Conference on Learning Representations, May 2024
  5. LETI: Learning to Generate from Textual Interactions
    Xingyao WangHao PengReyhaneh Jabbarvand, and Heng Ji
    In Findings of the Association for Computational Linguistics: (NAACL); DaSH Workshop @ NAACL 2024 (Oral), Jun 2024

2023

  1. Code4Struct: Code Generation for Few-Shot Event Structure Prediction
    Xingyao WangSha Li, and Heng Ji
    In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Jul 2023

awards


Outstanding Paper, NAACL 2024 2024
Yunni and Maxine Pao Memorial Fellowship 2024-2025
Honorable Mention, Computing Research Association Outstanding Undergraduate Researcher Award 2022
James B. Angell Scholar, University of Michigan 2021
EMNLP 2023 Student Travel Award (5 out of 94) 2021
University Honors, University of Michigan. 2017-2021

experiences


Google Research
Student Researcher
Summer 2023
New York City, NY, USA
Microsoft
Research Intern
Summer 2022
Remote, USA
ByteDance
Natural Language Processing Intern
Summer 2021
Beijing, China
Tencent
Software Engineering Intern
Summer 2020
Shenzhen, Guangdong, China