About
I am a Research Scientist in Alignment and Agents at Skywork AI, led by Prof Shuicheng Yan, working towards accelerating the realization of Artificial General Intelligence (AGI). I received my Ph.D. from Xidian University in 2021, supervised by Prof Bo Chen and Prof Mingyuan Zhou, and was a Research Fellow at Nanyang Technological University until 2024, supervised by Prof Bo An.
My research focuses on generative models (diffusion models and large language models), reinforcement learning from human feedback (RLHF), and their applications to AIGC, including multi-step reasoning, retrieval-augmented generation (RAG), and alignment in multimodal generation. I have also worked on Bayesian models, graph neural networks, self-supervised learning and out-of-distribution detection.
I am looking for research interns interested in LLMs for math reasoning and code generation. Please send me your CV if you are interested.
Ventures
News
- Serving as an Area Chair for NeurIPS 2026 in the area of Large Language Models.
- "Improving Multi-step Reasoning Abilities of LLMs with Direct Advantage Policy Optimization" accepted to NeurIPS 2025 Spotlight
- "Q*: Improving Multi-step Reasoning for LLMs with Deliberative Planning" released on arXiv.
- Joined Skywork AI as a Research Scientist in Alignment and Agents.
- Champion (1st place) of the L2RPN-ICAPS Learning to Run a Power Network Challenge.
Experience
- Skywork AIFeb 2024 – PresentResearch Scientist, Alignment and Agents · Singapore
- Nanyang Technological UniversityNov 2021 – Feb 2024Research Fellow, School of Computer Science and Engineering · Singapore
- Xidian University2016 – 2021Ph.D. in Signal and Information Processing · Xi'an, China
- Xidian University2012 – 2016B.E. in Electronic Engineering and Information Science · Xi'an, China
Awards & Services
- Champion (1st place), L2RPN-ICAPS Learning to Run a Power Network Challenge, 2021–2022
- Bronze Medal, ACM-ICPC China Shaanxi Provincial Programming Contest, 2016 (individual / team)
- Area Chair, NeurIPS 2026 (Large Language Models)
- Reviewer for AAAI 2021, ICLR 2021–2025, ICML 2021–2025, NeurIPS 2022–2025, IEEE TPAMI, IEEE TSP, IEEE TNNLS
Selected Publications
* equal contribution · # corresponding author · Full list on Google Scholar · BibTeX
Alignment & Agents
- Improving Multi-step Reasoning Abilities of Large Language Models with Direct Advantage Policy OptimizationJiacai Liu, Chaojie Wang#, Yang LiuNeurIPS 2025 Spotlight
- Q*: Improving Multi-step Reasoning for LLMs with Deliberative PlanningChaojie Wang*, Yancheng Deng*#, Zhiyi Lv, Shuicheng Yan, Bo AnTPAMI (under review), 2025arXiv
- Removing Length Bias in RLHF Is Not EnoughChaojie Wang, Haonan Shi, Long Tian#, Bo An, Shuicheng YanACL 2025
- MacroHFT: Memory Augmented Context-aware Reinforcement Learning on High Frequency TradingChuqiao Zong, Chaojie Wang#, Molei Qin, Lei Feng, Xinrun Wang, Bo AnKDD 2024
Generative Model Theory
- Candidate Label Set Pruning: A Data-centric Perspective for Deep Partial-label LearningShuo He, Chaojie Wang, Guowu Yang, Lei FengICLR 2024 Oral
- Few-shot Generation via Recalling the Episodic-Semantic Memory like Human BeingZhibin Duan*, Zhiyi Lv*, Chaojie Wang#, Bo Chen, Bo An, Mingyuan ZhouNeurIPS 2023
- Out-of-Distribution Detection with An Adaptive Likelihood Ratio on Informative Hierarchical VAEYewen Li*, Chaojie Wang*#, Xiaobo Xia, Tongliang Liu, Xin Miao, Bo AnNeurIPS 2022
- Alleviating "Posterior Collapse" in Deep Topic Models via Policy GradientYewen Li*, Chaojie Wang*#, Zhibin Duan, Dongsheng Wang, Bo Chen, Bo An, Mingyuan ZhouNeurIPS 2022
Natural Language Processing
- Generative Text Convolutional Neural Network for Hierarchical Document Representation LearningChaojie Wang, Bo Chen, Zhibin Duan, Wenchao Chen, Hao Zhang, Mingyuan ZhouIEEE TPAMI, 2022
- Learning Hierarchical Document Graphs From Multilevel Sentence RelationsHao Zhang*, Chaojie Wang*, Zhengjue Wang, Zhibin Duan, Bo Chen, Mingyuan Zhou, Ricardo Henao, Lawrence CarinIEEE TNNLS, 2022
- Bidirectional Convolutional Poisson Gamma Dynamical SystemsWenchao Chen*, Chaojie Wang*, Bo Chen, Yicheng Liu, Hao Zhang, Mingyuan ZhouNeurIPS 2020Paper
- Convolutional Poisson Gamma Belief NetworkChaojie Wang, Bo Chen, Sucheng Xiao, Mingyuan ZhouICML 2019PDFarXivCode
Graph Neural Networks
- A Variational Edge Partition Model for Supervised Graph Representation LearningYilin He*, Chaojie Wang*#, Hao Zhang, Bo Chen, Mingyuan Zhou#NeurIPS 2022
- Deep Relational Topic Modeling via Graph Poisson Gamma Belief NetworkChaojie Wang*, Hao Zhang*, Bo Chen, Dongsheng Wang, Zhengjue Wang, Mingyuan ZhouNeurIPS 2020Paper
- Learning Dynamic Hierarchical Topic Graph with Graph Convolutional Network for Document ClassificationZhengjue Wang*, Chaojie Wang*, Hao Zhang, Zhibin Duan, Bo Chen, Mingyuan ZhouAISTATS 2020