Xun Wang汪洵
Third-year Ph.D. student @ IIIS, Tsinghua.
Hi, I am Xun Wang!
I am currently a third-year Ph.D. student at Institute for Interdisciplinary Information Sciences (IIIS), Tsinghua University, where I am fortunate to be advised by Prof. Longbo Huang. Prior to this, I earned a Bachelor of Science degree in 2024 from the School of Computer Science at Nanjing University.
Research Interests
My research interests broadly lie in AI for decision-making, focusing on the algorithms and applications of Reinforcement Learning (RL), Large Language Models (LLMs), and their intersection. Currently, I am studying how to leverage frontier models and architectures to achieve reliable and efficient decision-making in complex settings.
News
- [2026.05] Four papers on RL and GFlowNets accepted to ICML 2026!
- [2026.02] Paper “OM²P: Offline Multi-Agent Mean-Flow Policy” accepted to AAMAS 2026!
- [2025.07] Paper “Beyond Static Populations: Efficient Delay-Constrained Scheduling for Dynamic Users via Deep Reinforcement Learning” accepted to MobiHoc 2025!
- [2025.07] Paper “Online Adaptable Offline RL with Guidance Model” accepted to TNNLS!
Preprints
arXiv2026
FlexLoop: Depth-Elastic Looped Policies for Adaptive Test-Time Computation in Deep RL
arXiv2026
When Context Returns: Toward Robust Internalization in On-Policy Distillation
arXiv2025
From Solo to Symphony: Orchestrating Multi-Agent Collaboration with Single-Agent Demos
arXiv2026
The Router Within: Eliciting Native Skill Routing from a Frozen LLM
arXiv2026
Optimal Skill Selection for LLM Agents with Provable Bicriteria Guarantees
arXiv2026
MLREF: Efficient Module Reuse for Reward Design in Reinforcement Learning via Large Language Models
Selected Publications
* indicates equal contribution ConferenceJournal
ICML2026
Beyond the Proxy: Trajectory-Distilled Guidance for Offline GFlowNet Training
ICML2026
Reparameterization Proximal Policy Optimization
ICML2026
Reparameterization Flow Policy Optimization
ICML2026
Diffusing to Coordinate: Efficient Online Multi-Agent Diffusion Policies
QCNC2026
Effective Entanglement Scheduling in Quantum Networks via Deep Reinforcement Learning
MobiHoc2025
TNNLS2025
Online Adaptable Offline RL with Guidance Model
