PeopleofSuperintelligence

Lab leader

Yang Su

Yang Su works on language-model agents, from virtual-reality interaction to agent-world models. He interned on graph machine learning at AWS AI Lab, introduced Voice2Action, and helped build ALICE for generating data through agent interaction. He joined Qwen Team as a researcher in September 2024 and coauthored technical reports on Qwen2.5 and Qwen3. His 2026 work includes Qwen-AgentWorld and E-Commerce Bench; as of October 2026, he leads fundamental agent research at Qwen Team.

Now at
Cornell ICPC, Millennium and Qwen Team

Timeline

  1. 2021–2022

    Graph learning at AWS

    He interns on graph machine learning at AWS AI Lab in 2021–2022 and contributes to the open-source Deep Graph Library.

  2. 2023

    Voice commands in virtual reality

    In 2023, he introduces Voice2Action, a framework that extracts actions and entities from voice and text commands for real-time interaction in virtual reality.

  3. Since 2024

    Building ALICE

    From January 2024, he works part-time with Cornell ICPC and Millennium on ALICE, a meta-agent system that generates data through multi-turn interactions and feedback.

  4. September 2024

    Joining the Qwen Team

    He joins Qwen Team as a researcher in September 2024.

  5. December 2024

    Qwen2.5 technical report

    In December 2024, he coauthors the Qwen2.5 Technical Report. It describes post-training with more than one million supervised fine-tuning samples and multistage reinforcement learning.

  6. May 2025

    Qwen3 technical report

    In May 2025, he coauthors the Qwen3 Technical Report, which describes dense and Mixture-of-Experts language models with scales from 0.6 billion to 235 billion parameters.

  7. June 2026–August 2026

    World models for agents

    In June 2026, he coauthors Qwen-AgentWorld, a language world model for simulating agent environments across seven domains. In August 2026, he coauthors E-Commerce Bench, a benchmark for long-horizon business operations by LLM agents.

  8. As of October 2026

    Leading agent research

    As of October 2026, he leads fundamental agent research at Qwen Team. His interests include long-horizon reinforcement learning for agent alignment.

Sources and credits

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

  1. yang-su2000 (Yang Su) · GitHub github.com. Checked 15 October 2026.
  2. [2310.00092] Voice2Action: Language Models as Agent for Efficient Real-Time Interaction in Virtual Reality arxiv.org. Checked 15 October 2026.
  3. GitHub - yang-su2000/Voice2Action: ALICE and its prior work, Voice2Action: Language Models as Agent for Efficient Real-Time Interaction in Virtual Reality github.com. Checked 15 October 2026.
  4. [2412.15115] Qwen2.5 Technical Report arxiv.org. Checked 15 October 2026.
  5. [2505.09388] Qwen3 Technical Report arxiv.org. Checked 15 October 2026.
  6. [2606.24597] Qwen-AgentWorld: Language World Models for General Agents arxiv.org. Checked 15 October 2026.
  7. [2608.30730] E-Commerce Bench: Evaluating LLM Agents on Long-Horizon Autonomous Business Operation arxiv.org. Checked 15 October 2026.
  8. yang-su2000 (Yang Su) huggingface.co. Checked 15 October 2026.