Yibo Peng彭艺博

Péng Yìbó (pronounced "penɡ EE-bor")

Carnegie Mellon University

Pittsburgh, Pennsylvania
Yibo Peng

About

I am Yibo Peng, a Research Assistant at InfiniAI Lab @ CMU , advised by Prof. Beidi Chen. My research centers on AI security in real-world applications. Currently, I am also an incoming visiting student working with Prof. David Wagner (UC Berkeley) and PhD student Sizhe Chen. Additionally, I collaborate closely with Prof. Huaxiu Yao on multimodal large language models.

Previously, I earned my Master's degree in Artificial Intelligence Engineering from CMU. During my graduate studies, I was fortunate to work with Prof. Daniel Fried and phd student Zora Zhiruo Wang on Code Generation, and with Prof. Graham Neubig on code agents.

Broadly, my research aims to address core challenges for AI agents: enhancing multimodal reasoning, optimizing collaboration, and ensuring safety and reliability.

🎓 I am actively seeking Ph.D. opportunities for Fall 2026. If you are interested in my work or potential collaborations, please feel free to contact me at [email protected].

News

[Dec 2025] 🔥 I will attend NeurIPS 2025 in San Diego, CA. Excited to catch up with old and new friends!
[Jun 2025] 🔥 I will be joining Prof. Beidi Chen's InfiniAI Lab @ CMU.

Selected Publications

Simple OCR
Context-Contrastive Reinforcement Learning for Robust MLLM Reading
Yibo Peng*, Peng Xia*, Ding Zhong*, Kaide Zeng, Siwei Han, Yiyang Zhou, Jiaqi Liu, Ruiyi Zhang, Huaxiu Yao
Submitted to ACL 2026 (ARR Jan)
Code Agents Safety
When "Correct" Is Not Safe: Can We Trust Functionally Correct Patches Generated by Code Agents?
Yibo Peng*, James Song*, Lei Li*, Xinyu Yang, Mihai Christodorescu, Ravi Mangal, Corina Pasareanu, Haizhong Zheng, Beidi Chen
Submitted to ACL 2026 (ARR Oct, Meta Review: 4)
Accepted by ResponsibleFM @ NeurIPS 2025
MMedAgent-RL
MMedAgent-RL: Optimizing Multi-Agent Collaboration for Multimodal Medical Reasoning
Peng Xia*, Yibo Peng*, Jinglu Wang*, Kaide Zeng, Xian Wu, Xiangru Tang, Hongtu Zhu, Yun Li, Yan Lu, Huaxiu Yao
Submitted to ICLR 2026 (Average Rating: 6)
Long-context Code Generation
Can Long-Context Language Models Solve Repository-Level Code Generation?
Yibo Peng, Zora Zhiruo Wang, Daniel Fried
arXiv preprint, 2025

Services

Reviewer: CVPR 2026, NeurIPS 2025 workshop, ICML 2025 workshop