Duo Wu

RECENT NEWS

RESEARCH INTERESTS

My research centers on decision intelligence, exploring how reinforcement learning (RL) can unlock stronger generalization, adaptation, and decision-making capabilities in pretrained generalist models.

My previous research has explored these questions in network intelligence and LLM-based autonomous agents. More recently, I am especially interested in embodied intelligence, seeking to understand how general reward systems can arise from interaction and provide scalable learning signals for generalist robot policies, enabling them to learn beyond demonstrations and tackle long-horizon, precise-control tasks.

PUBLICATION

Note: Equal contributions are marked by *. Advised students are underlined.

Selected Publications:

• Demonstration-Free Success-Probability Reward Learning for Generalist Robot Policies

  In Submission [CCF-A] preprint code

  Duo Wu, Haifeng Wang, Rongwei Lu, Jinghe Wang, Tianyi Xiong, Zhimin Wang, Chao Yu, Shuai Ma, Zhi Wang.

  We show that sparse task outcomes are sufficient to learn dense success-probability rewards from policy experience, enabling demonstration-free reward learning and evolving reward feedback for generalist robot policies.

• Trailblazer: Grounding Large Language Models as Policies for Autonomous Network Control

  In Submission to IEEE Transactions on Mobile Computing [CCF-A] preprint code

  Duo Wu, Linjia Kang, Zhimin Wang, Fangxin Wang, Wei Zhang, Chongbo Sun, Xuefeng Tao, Wei Yang, Le Zhang, Peng Cui, Wenwu Zhu, Zhi Wang.

  We present Trailblazer (开拓者), which represents a new control paradigm leveraging LLMs as generalizable policies to achieve unprecedented generalization across diverse tasks and environments.

• CATP-LLM: Empowering Large Language Models for Cost-Aware Tool Planning

  Accepted by ICCV 2025 [CCF-A] paper code

  Duo Wu*, Jinghe Wang*, Yuan Meng*, Yanning Zhang, Le Sun, Zhi Wang.

  We propose CATP-LLM that uses reinforcement learning to reduce costs of LLM tool use without sacrificing performance. We also establish the first platform OpenCATP to systematically evaluate the effectiveness of LLMs in cost-aware tool use.

• NetLLM: Adapting Large Language Models for Networking

  Accepted by ACM SIGCOMM, 2024 [CCF-A, only 23 papers from mainland China were accepted] paper code slides

  Duo Wu, Xianda Wang, Yaqi Qiao, Zhi Wang, Junchen Jiang, Shuguang Cui, Fangxin Wang.

  NetLLM is the first systematic transfer learning framework that adapts LLMs to solve decision making problems in networking. It provides valuable insights on LLM domain adaptation for the broad research communities. As of 2026/09, it has accumulated 300+ citations and 200+ Github stars.

Full List of Publications:

Accepted:

1. GeoMIND: A Benchmark for Spatial Understanding in Robotic Manipulation

   Accepted by NeurIPS 2026[CCF-A]

   Jinghe Wang*, Xinrui Cao*, Duo Wu*, Chenghao Gu, Yong Zhong, Linjia Kang, Tianyi Xion, Zhi Wang.

2. DeCoPatch: Revealing Causal Latent Subspaces in Vision-Language Models for GUI Grounding

   Accepted by ECCV 2026 [THU-A] preprint webpage

   Yongkang Zhang*, Linjia Kang*, Zhimin Wang*, Duo Wu, Zhi Wang.

3. IGen: Scalable Data Generation for Robot Learning from Open-World Images

   Accepted by CVPR 2026 [CCF-A] preprint code webpage

   Chenghao Gu*, Haolan Kang*, Junchao Lin*, Jinghe Wang, Duo Wu, Shuzhao Xie, Fanding Huang, Junchen Ge, Ziyang Gong, Letian Li, Hongying Zheng, Changwei Lv, Zhi Wang.

4. CATP-LLM: Empowering Large Language Models for Cost-Aware Tool Planning

   Accepted by ICCV 2025 [CCF-A] paper code

   Duo Wu*, Jinghe Wang*, Yuan Meng*, Yanning Zhang, Le Sun, Zhi Wang.

5. Cluster Based Heterogeneous Federated Foundation Model Adaptation and Fine-Tuning

   Accepted by AAAI, 2025 [CCF-A] paper

   Xianda Wang*, Yaqi Qiao*, Duo Wu*, Chenrui Wu, Fangxin Wang.

6. NetLLM: Adapting Large Language Models for Networking

   Accepted by ACM SIGCOMM, 2024 [CCF-A] paper code slides

   Duo Wu, Xianda Wang, Yaqi Qiao, Zhi Wang, Junchen Jiang, Shuguang Cui, Fangxin Wang.

7. MANSY: Generalizing Neural Adaptive Immersive Video Streaming With Ensemble and Representation Learning

   Accepted by IEEE Transactions on Mobile Computing (TMC), 2024 [CCF-A] paper code

   Duo Wu, Panlong Wu, Miao Zhang, Fangxin Wang.

8. ILCAS: Imitation Learning-Based Configuration-Adaptive Streaming for Live Video Analytics with Cross-Camera Collaboration

   Accepted by IEEE Transactions on Mobile Computing (TMC), 2023 [CCF-A] paper

   Duo Wu, Dayou Zhang, Miao Zhang, Ruoyu Zhang, Fangxin Wang, Shuguang Cui.

9. A Comprehensive Survey on Segment Routing Traffic Engineering

   Accepted by Digital Communications and Networks (DCN), 2022 [JCR-Q1] paper

   Duo Wu, Lin Cui.

In Submission:

1. Demonstration-Free Success-Probability Reward Learning for Generalist Robot Policies

   In Submission [CCF-A] preprint code webpage

   Duo Wu, Haifeng Wang, Rongwei Lu, Jinghe Wang, Tianyi Xiong, Zhimin Wang, Chao Yu, Shuai Ma, Zhi Wang.

2. Trailblazer: Grounding Large Language Models as Generalizale Policies for Network Control

   In Submission to IEEE Transactions on Mobile Computing [CCF-A] preprint code

   Duo Wu, Linjia Kang, Zhimin Wang, Fangxin Wang, Wei Zhang, Chongbo Sun, Xuefeng Tao, Wei Yang, Le Zhang, Peng Cui, Wenwu Zhu, Zhi Wang.

3. Learning with Challenges: Adaptive Difficulty-Aware Data Generation for Mobile GUI Agent Training

   In Submission [CCF-A] preprint webpage

   Linjia Kang*, Zhimin Wang*, Yongkang Zhang*, Duo Wu, Jinghe Wang, Ming Ma, Haopeng Yan, Zhi Wang.

4. Collaborative Belief Reasoning with LLMs for Efficient Multi-Agent Collaboration

   In Submission [CCF-A] preprint

   Zhimin Wang*, Duo Wu*, Shaokang He*, Linjia Kang, Jinghe Wang, Jing Yu, Kai Zhu, Jiawei Li, Zhi Wang.

STANDARD

• Information Technology Digital Retina Systems Part 4: Edge Subsystem

  Artificial Intelligence Technology Industry Strategic Alliance (AITISA), T/AI 116.4-2025

  Duo Wu, Yaowei Wang, Yuan Xue, Wen Ji, Ying Wang, Qingfang Zheng, Xinbei Bai, Zhi Wang, Peng Chen, Le Sun, Yunhong Zhou, Qiben Shan, Jiajun Luo, Jiacheng Jiang, Chen Tang, Yan Lan, Pan Li, Jinyu Yuan, Weisheng Kong, Xiaolin Yang, Changyu Liu, Haijun Liu, Xue Rao, Jiangang Zhou, Rongwei Lu, Shuzhao Xie, Yuan Meng, Xuanti Liu, Wenwu Zhu, Wen Gao.

EXPERIENCE

• Network Transmission Research Group, Bytedance, Shenzhen

  Research Intern, from Jan. 2025 to Oct. 2025.

• Future Network of Intelligence Institute, CUHKSZ

  Research Assistant, from Sep. 2022 to April 2024.

• Huawei Technologies Co., Ltd., Dongguan

  Software Engineer Intern, from July. 2021 to Sep. 2021.

AWARD

• ICML Gold Reviewer, 2026

• CIC Master's Thesis Incentive Program, 2025

  Only 7 awardee in the nation.

• Presidential Award for Outstanding Graduate Students, CUHKSZ, 2024

  Only 10 awardee out of 1500+ postgraduate students in CUHKSZ

• Academic Star Nomination, JNU, 2021

  Only 23 awardee in JNU

• University Scholarship of Innovative and Talented Undergraduate, JNU, 2021

  10000 RMB, only 30 awardee in JNU