👨🎓 About Me
I am studying Computer Science and Technology at the College of Computer Science and Technology, National University of Defense Technology (NUDT), under the supervision of Prof. Wenjing Yang and Prof. Shixuan Liu.
My research focuses on reliable reasoning for foundation models, with a particular emphasis on reasoning verification, neuro-symbolic reasoning, and trustworthy AI. I study how to identify, verify, and improve failures in multi-step reasoning systems.
I am especially interested in combining formal or probabilistic reasoning with modern language models, and in developing interpretable signals for understanding when and why reasoning fails.
My research interests include Reasoning Verification, Neuro-Symbolic AI, Trustworthy Foundation Models, and Formal & Probabilistic Reasoning.
I also serve as a reviewer for ICLR.
✨ News
- 2026.09: 🎉 Mining Logic under Uncertainty was accepted by NeurIPS 2026.
- 2026.04: 🎉 LogicSAGE was accepted by ICML 2026.
- 2026.01: 🏆 Logical-SAGE received an Outstanding Paper Award at the AAAI 2026 Bridge on Logical and Symbolic Reasoning in Language Models.
📝 Publications
* Equal contribution.
Mining Logic under Uncertainty: Probabilistic Soft Logic with Energy-Based Inference for Chain-of-Thought Verification
Jiang Yu, Jinlong Tian, Kewei Cheng, Yue He, Haoxuan Li, Haotian Wang, Yunhai Wang, Wenjing Yang, Zhouchen Lin, Shixuan Liu
NeurIPS 2026
LogicSAGE: Neuro-Symbolic Reasoning with Socratic-Guided Enhancement
Jinlong Tian*, Jiang Yu*, Kewei Cheng, Fengxiang Cheng, Yue He, Yunfei Wang, Haotian Wang, Haoxuan Li, Wenjing Yang, Shixuan Liu
ICML 2026
Logical-SAGE: A Logical Socratic Architecture for Guided Evolution in Neuro-Symbolic Reasoning
Jinlong Tian*, Jiang Yu*, Kewei Cheng, Yue He, Yunfei Wang, Huibin Tan, Haotian Wang, Wenjing Yang, Shixuan Liu
AAAI 2026 Bridge: Logical and Symbolic Reasoning in Language Models · Outstanding Paper Award
🔬 Research
- Reasoning Verification. Auditing multi-step reasoning and checking whether conclusions are supported by valid reasoning processes.
- Neuro-Symbolic Reasoning. Combining foundation models with structured, formal, and probabilistic reasoning mechanisms.
- Reasoning Dynamics. Studying internal representations and signals associated with reasoning failures and change points.
- Trustworthy AI. Building verification and diagnostic tools that make foundation-model behavior more reliable and interpretable.