I am a Ph.D. candidate in Peking University, supervised by Prof. Ge Li. I completed my undergraduate studies in School of Computer Science in Wuhan University, where I ranked 1st out of 57. I have also been fortunate to begin my research journey under the supervision of Prof. Mang Ye in 2024.

My research interest includes code generation, formal verification and LLM reasoning. I aspire to work on research that introduces genuinely new paradigms and challenges prevailing assumptions in both academia and industry.

🔥 News

• 2026.07:  🎉🎉 Paper Saber has been awarded ACL 2026 SAC Highlight Award.
• 2026.03:  🎉🎉 Paper KOCO-Bench has been accepted by ACL Main Conference 2026.
• 2025.09:  🎉🎉 Our survey paper has been accepted by Software Journal 2025.
• 2025.09:  🎉🎉 Paper OASIS has been accepted by NeurIPS 2025.
• 2025.05:  🎉🎉 Paper GHOST has been accepted by ICML 2025.

📝 Selected Publications

NeurIPS 2026

From I/O to Code with Discovery Agent

Yihong Dong, Jiaru Qian, Haoran Zhang, Peixu Wang, Binhua Li, Zhi Jin, Yongbin Li, Ge Li, Xiaokang Yang, Xue Jiang

Introduces IO2Code and a discovery agent that synthesizes programs from input-output examples through curriculum evolution.

ACL 2026

KOCO-Bench: Can Large Language Models Leverage Domain Knowledge in Software Development?

Xue Jiang, Ge Li, Jiaru Qian, Xianjie Shi, Chenjie Li, Hao Zhu, Ziyu Wang, Jielun Zhang, Zeyu Zhao, Kechi Zhang, Jia Li, Wenpin Jiao, Zhi Jin, Yihong Dong

Benchmarks how well language models acquire and apply new domain knowledge in realistic software development.

Software Journal 2025

A Survey on Code Generation with LLM-based Agents

Yihong Dong*, Xue Jiang*, Jiaru Qian*, Tian Wang, Kechi Zhang, Zhi Jin, Ge Li

Provides a systematic taxonomy of LLM-based code-generation agents and outlines key challenges and future directions.

NeurIPS 2025

OASIS: One-Shot Federated Graph Learning via Wasserstein Assisted Knowledge Integration

Frank Wan*, Jiaru Qian*, Wenke Huang, Qilin Xu, Xianda Guo, Boheng Li, Guibin Zhang, Bo Du, Mang Ye

Integrates semantic and structural knowledge with a topology codebook and Wasserstein-guided distillation.

ICML 2025

GHOST: Generalizable One-Shot Federated Graph Learning with Proxy-Based Topology Knowledge Retention

Jiaru Qian*, Guancheng Wan*, Wenke Huang*, Guibin Zhang*, Yuxin Wu, Bo Du, Mang Ye

Retains transferable topology knowledge through learnable proxies for generalizable one-shot federated graph learning.

Under Review

Self-Spec Verifiable Code Generation

Jiaru Qian · First Author

Benchmarks the capability of LLMs on formally verifiable code generation.

Under Review

Parallel Thinking for Language Model with Path Diversity

Jiaru Qian · Co-First Author

Encourages diverse reasoning paths during reinforcement learning to prevent mode collapse in parallel thinking.

* Equal Contribution

🎖 Honors and Awards

• 2026.06 Outstanding Graduate of Wuhan University
• 2026.05 Lei Jun Computer Science Breakthrough Scholarship, Wuhan University (CNY 50,000)
• 2022–2026 National Scholarship, awarded for three consecutive years (CNY 10,000 each)
• 2024.12 National Second Prize, China Undergraduate Mathematical Contest in Modeling

📖 Education

• 2026–2031, Ph.D. in Computer Science, School of Computer Science, Peking University
• 2022–2026, B.Eng. in Computer Science, School of Computer Science, Wuhan University