Wen-Chao Hu (胡文超)

Wen-Chao Hu 胡文超 Ph.D. Candidate, LAMDA Group

I am currently pursuing my Ph.D. degree in LAMDA Group, Nanjing University, where I have the great privilege of being advised by Prof. Yuan Jiang and Prof. Zhi-Hua Zhou. Previously, I received my B.Eng. degree in June 2021 from School of Intelligent Systems Engineering, Sun Yat-sen University.

My research focuses on advancing reliable 🔧 and efficient ⚡ AI methods, particularly in:

Abductive Learning and Neuro-Symbolic Methods: Integrating the power of data-driven AI (e.g., machine learning, LLMs) and knowledge-driven AI (e.g., logical reasoning) to build AI systems that are logically reliable and suitable for high-stakes applications.
Optimization: Developing efficient algorithms for joint optimization involving both continuous numerical variables and discrete symbolic structures, a core challenge in computer science (including but not limited to neuro-symbolic integration).

We hold a weekly Abductive Learning (ABL) Seminar. If you are interested in joining, feel free to contact me.

Besides research, I am passionate about strength training, CrossFit 🏋️‍♂️ and racing bicycle 🚴‍♂️, see photos.

CV 📄 / Email ✉️

Photo taken by Yi Shi in Philadelphia, 2025.

📖 Publication

Conference Paper

Curriculum Abductive Learning. [PDF, arXiv, poster, bibtex]
Wen-Chao Hu, Qi-Jie Li, Lin-Han Jia, Cunjing Ge, Yu-Feng Li, Yuan Jiang, and Zhi-Hua Zhou.
In: Advances in Neural Information Processing Systems 38 (NeurIPS 2025). Page: to appear. CCF-A

We enhance ABL efficiency, stability, and reasoning accuracy by controlling the knowledge base complexity throughout training.
```
@inproceedings{hu2025cabl,
  title={Curriculum Abductive Learning},
  author={Hu, Wen-Chao and Li, Qi-Jie and Jia, Lin-Han and Ge, Cunjing and Li, Yu-Feng and Jiang, Yuan and Zhou, Zhi-Hua},
  booktitle = {Advances in Neural Information Processing Systems 38 (NeurIPS)},
  year={2025},
  pages = {to appear}
}
```

Verification Learning: Make Unsupervised Neuro-Symbolic System Feasible. [PDF, arXiv, bibtex]
Lin-Han Jia, Wen-Chao Hu, Jie-Jing Shao, Lan-Zhe Guo, and Yu-Feng Li.
In: Proceedings of the 42nd International Conference on Machine Learning (ICML 2025). Page: 27292-27306. CCF-A

We propose a label-free neuro-symbolic method by verification, formulated as a constraint optimization problem and solved via DCS.
```
@inproceedings{jia2025vl,
  title={Verification Learning: Make Unsupervised Neuro-Symbolic System Feasible},
  author={Jia, Lin-Han and Hu, Wen-Chao and Shao, Jie-Jing and Guo, Lan-Zhe and Li, Yu-Feng},
  booktitle = {Proceedings of the 42nd International Conference on Machine Learning (ICML)},
  year = {2024},
  pages = {27292-27306}
}
```

Efficient Rectification of Neuro-Symbolic Reasoning Inconsistencies by Abductive Reflection. [PDF, arXiv, poster, slides, code, bibtex]
Wen-Chao Hu, Wang-Zhou Dai, Yuan Jiang, and Zhi-Hua Zhou.
In: Proceedings of the 39th AAAI Conference on Artificial Intelligence (AAAI 2025 Outstanding Paper Award). Page: 17333-17341. CCF-A

♣ Extended abstract version [PDF], published in: Proceedings of the 34th International Joint Conference on Artificial Intelligence
(IJCAI 2025 Best Papers from Sister Conferences Track). Page: 10896-10900. CCF-A

We propose an efficient ABL extension combining neural networks and symbolic reasoning with both sides' integrity preserved.
```
@inproceedings{hu2025reflection,
  title      = {Efficient Rectification of Neuro-Symbolic Reasoning Inconsistencies by Abductive Reflection},
  author     = {Hu, Wen-Chao and Dai, Wang-Zhou and Jiang, Yuan and Zhou, Zhi-Hua},
  booktitle  = {Proceedings of the 39th AAAI Conference on Artificial Intelligence (AAAI)},
  pages      = {17333-17341},
  year       = {2025}
}
```

Knowledge-Enhanced Historical Document Segmentation and Recognition. [PDF, bibtex]
En-Hao Gao, Yu-Xuan Huang, Wen-Chao Hu, Xin-Hao Zhu, and Wang-Zhou Dai.
In: Proceedings of the 38th AAAI Conference on Artificial Intelligence (AAAI 2024 Oral). Page: 8409--8416. CCF-A

We apply ABL to enhance historical document OCR through effective use of formal expert knowledge.
```
@inproceedings{gao2024kesar,
  title      = {Knowledge-Enhanced Historical Document Segmentation and Recognition},
  author     = {Gao, En-Hao and Huang, Yu-Xuan and Hu, Wen-Chao and Zhu, Xin-Hao and Dai, Wang-Zhou},
  booktitle  = {Proceedings of the 38th AAAI Conference on Artificial Intelligence (AAAI)},
  pages      = {8409--8416},
  year       = {2024}
}
```

Journal Paper

ABLkit: A Python Toolkit for Abductive Learning. [PDF, bibtex]
Yu-Xuan Huang, Wen-Chao Hu, En-Hao Gao, and Yuan Jiang.
Frontiers of Computer Science, 2024, 18(6):186354. CCF-B

@article{huang2024ablkit,
  title      = {ABLkit: a Python toolkit for abductive learning},
  author     = {Huang, Yu-Xuan and Hu, Wen-Chao and Gao, En-Hao and Jiang, Yuan},
  journal    = {Frontiers of Computer Science},
  volume     = {18},
  number     = {6},
  pages      = {186354},
  year       = {2024}
}

Preprint

When Is Prior Knowledge Helpful? Exploring the Evaluation and Selection of Unsupervised Pretext Tasks from a Neuro-Symbolic Perspective. [arXiv, bibtex]
Lin-Han Jia, Si-Yu Han, Wen-Chao Hu, Jie-Jing Shao, Wen-Da Wei, Zhi Zhou, Lan-Zhe Guo, and Yu-Feng Li.

We unify the theoretical frameworks of NeSy and SSL, linking pretext task success to knowledge learnability, reliability, and completeness.
```
@article{jia2025prior,
  title={When Is Prior Knowledge Helpful? Exploring the Evaluation and Selection of Unsupervised Pretext Tasks from a Neuro-Symbolic Perspective},
  author={Jia, Lin-Han and Han, Si-Yu and Hu, Wen-Chao and Shao, Jie-Jing and Wei, Wen-Da and Zhou, Zhi and Guo, Lan-Zhe and Li, Yu-Feng},
  journal={arXiv preprint arXiv:2508.07299},
  year={2025}
}
```

🧪 Projects

ABLkit [GitHub, PyPI]: I am the core developer and current maintainer of ABLkit, an efficient open-source toolkit for abductive learning, which leverages the power of both data and knowledge. For more information, please refer to: ABLkit Documentation.

🎉 Awards

National Scholarship for Doctoral Students, 2025. (¥30,000 award)
NeurIPS Scholar Award, 2025.
AAAI Outstanding Paper Award, 2025. (3 outstanding papers out of 12,957 submissions) [certificate, ceremony, 中文采访]
Second Prize (3rd place), Greater Bay International Algorithm Case Competition, 2022. (¥100,000 award) [certificate, ceremony]
Outstanding Bachelor's Thesis of Sun Yat-sen University, 2021. (Thesis Title: Robust Principal Component Analysis)
National Scholarship for Undergraduate Students, 2018, 2019.

🎓 Teaching Assistant

Digital Signal Processing (with Prof. Wei Wang; 2022 fall, 2023 fall, 2025 spring, 2025 fall)
Introduction to Machine Learning (with Prof. Zhi-Hua Zhou and Assoc. Prof. Peng Zhao; 2024 spring)
Mathematical Logic (with Dr. Cunjing Ge; 2023 spring, 2024 spring)
Introduction to Machine Learning (with Prof. Zhi-Hua Zhou and Assoc. Prof. Han-Jia Ye; 2022 fall)
Symbolic Learning (with Assoc. Prof. Wang-Zhou Dai; 2022 spring)
Operating Systems (with Dr. Xintao Niu and Huayao Wu; 2022 spring)

🧰 Academic Service

PC Member or Reviewer for Conferences: AISTATS (2025, 2026), CVPR (2026), ICLR (2025, 2026), ICML (2025, 2026), IJCAI (2025), IJCLR (2024, 2025), NeSy (2024, 2025), NeurIPS (2024, 2025), PAKDD (2026), LMReasoning@AAAI (2026).
Volunteer for Conferences: MLA (2023), IJCLR (2024).
Server Administrator for LAMDA group, 2021-2022.

🌐 Correspondence

Email:
huwc@lamda.nju.edu.cn

Office:
Room 912, Computer Science Building, Xianlin Campus of Nanjing University

Mail Address:
National Key Laboratory for Novel Software Technology, Nanjing University, Xianlin Campus, 163 Xianlin Avenue, Qixia District, Nanjing 210023, China
(江苏省南京市栖霞区仙林大道163号, 南京大学仙林校区, 软件新技术国家重点实验室, 210023)