colin@world:~ $ whoami

[colin@world ~]$ whoami

[colin@world ~]$ cat interests.md

My current research focuses on code reasoning models:
- Current LLMs still struggle with understanding real code semantics beyond literal features. I am interested in how to learn those semantics better. Additionally, learning code semantics can be seen as a loose abstraction of other realistic reasoning tasks, so it is interesting to see how code reasoning can benefit others.
- "World modeling" for code is expected to improve code semantics understanding, but current results show limited benefits of transferability, which motivates me to explore "why" and "how".

[colin@world ~]$ cat ./pubs/*_selected*

[LLM4Code '25] CWEval: Outcome-driven Evaluation on Functionality and Security of LLM Code Generation
Jinjun Peng, Leyi Cui, Kele Huang, Junfeng Yang, Baishakhi Ray
[arXiv:2501.08200] [Code]
[NeurIPS '24] SemCoder: Training Code Language Models with Comprehensive Semantics Reasoning
Yangruibo Ding, Jinjun Peng, Marcus J Min, Gail Kaiser, Junfeng Yang, Baishakhi Ray
[arXiv:2406.01006] [Code]
[ESEC/FSE '23] NeuRI: Diversifying DNN Generation via Inductive Rule Inference
Jiawei Liu, Jinjun Peng, Yuyao Wang, and Lingming Zhang
[arXiv:2302.02261] [Code 0] [Code 1]
[OOPSLA '24] Quarl: A Learning-Based Quantum Circuit Optimizer
Zikun Li, Jinjun Peng, Yixuan Mei, Sina Lin, Yi Wu, Oded Padon, Zhihao Jia
[arXiv:2307.10120] [Code]

[colin@world ~]$ cat ./exps/*

Amazon, Applied Scientist Intern, 2025.05 - 2025.08
- Agentic retrieval for coding tasks
Columbia University, Ph.D. student in Computer Science, 2023.09 - ?
- Advised by Professors Baishakhi Ray and Junfeng Yang.
Tsinghua University, B.E. in Computer Science, 2019 - 2023
Beijing National Day School, 2016 - 2019
- Where dreams begin.

[colin@world ~]$ cat services.md

Program Committee/Reviewer
- ICLR ['26]
- DL4C @ NeurIPS ['25], ASE NIER '25
- TSE ['25], TOSEM ['25], TKDD ['25], Journal of Big Data

[colin@world ~]$ cat ack.md

[colin@world ~]$ curl slow-science.org

Jinjun Peng