Junjie Wang (王军杰)

Junjie Wang (王军杰)

Postdoctoral Fellow

About me

Junjie Wang avatar

Junjie Wang (王军杰) is currently a postdoctoral researcher at the IIGROUP Lab, Tsinghua University, supervised by Prof. Yujiu Yang. His research focuses on Learning to Reason for Self-Evolving Agents: enabling agents to learn from knowledge, digital interaction, and physical-world consequences, and ultimately to improve not only their capabilities but also the mechanisms by which they improve. He received his Ph.D. in Engineering from Waseda University, under the supervision of Prof. Tetsuya Sakai.

Email: wangjunjie@sz.tsinghua.edu.cn (Please state your purpose / 请您注明来意)

📰 News / 最近

  • [08*2026] 📜 8 papers are accepted to EMNLP 2026 !
  • [06/2026] 🏅 Selected as an ICML 2026 Golden Reviewer.
  • [04/2026] 📜 1 paper is accepted to ICML 2026 !
  • [04/2026] 📜 2 papers are accepted to ACL 2026 !
  • [03/2026] 📜 1 paper is accepted to TMLR ! 获选CAAI-腾讯犀牛鸟研究计划 !
  • [01/2026] 🎉 ROCO@AAAI2026 决赛冠军: URL
  • [01/2026] 📜 1 paper is accepted to ICLR 2026 !
  • [11/2025] 📜 1 paper (Oral) is accepted to AAAI 2026 !

🌌 Research Vision / 研究愿景

Learning to Reason: Toward Reasoning-Centric Recursive Self-Improvement

学会推理：迈向以推理为中心的递归自我改进

My central question is: How can reasoning become a continually improving capability? I study how agents learn to reason from knowledge, digital interaction, and physical-world consequences as human predefinition decreases and agent autonomy increases.

我的核心问题是：如何让推理成为一种能够持续改进的能力？ 我关注智能体如何从知识、数字交互和物理世界后果中学习推理，并随着人类预定义减少而逐步提升自主性。

Three stages from knowledge-grounded reasoning to interaction-grounded and physical-world reasoning

From human-defined knowledge to interaction feedback and real-world consequences.
从人类预定义知识，走向交互反馈与真实世界后果。

🧭 Three Reasoning Layers / 三层推理

  1. Knowledge-Grounded Reasoning / 知识驱动推理 — Learn representations and explicit reasoning structures from human-curated data, evidence, and tasks.
    从人类定义的数据、证据与任务中学习表征和显式推理结构。

  2. Interaction-Grounded Reasoning / 交互驱动推理 — Diagnose trajectories in tools, code, and GUIs, then convert verified feedback into capability updates.
    诊断工具、代码与 GUI 中的交互轨迹，并将可验证反馈转化为能力更新。

  3. Physical-World Reasoning / 物理世界推理 — Reason about dynamics, actions, consequences, and safety in robotics and autonomous driving.
    在机器人与自动驾驶中理解动力学、行动、后果与安全约束。

🔄 Toward Reasoning-Centric RSI / 迈向以推理为中心的 RSI

Interaction → Verification → Attribution → Internalization → Capability Improvement ↺

My long-term goal is Reasoning-Centric Recursive Self-Improvement (RSI): agents that can verify which experiences are trustworthy, attribute success and failure, internalize high-value lessons, and eventually improve the mechanisms by which they improve.

我的长期目标是构建以推理为中心的递归自我改进（Reasoning-Centric RSI）：让智能体能够验证经验、归因成败、内化高价值反馈，并最终改进其自身的改进机制。

✒️ Recent Professional Services / 最近的专业服务

  • Golden Reviewer: ICML 2026
  • Area Chair: ACL ARR 2025 October (EACL); ACL ARR 2026 January (ACL); ACL ARR 2026 March (EMNLP).
  • Senior PC member: WSDM 2026
  • Workshop Organiser: SIGIR-AP 2025 BREV-RAG Workshop

Publications

Details in Publications Page

Total Publications: 37

  • ⭐: Co-first Author
  • 🚩: Corresponding Author
  • 💭: Under Review