NEW BRUNSWICK, NJHello, I am Zeru Shi, a first-year Ph.D. student in Computer Science at Rutgers University. I am honored to be guided by Prof. Ruixiang(Ryan) Tang. I received my B.E. degree in Software Engineering at Dalian University of Technology.
My research interests mainly focus on:
I am highly interested in post-training for LLMs and VLMs. My focus is on developing more effective reinforcement learning algorithms, as well as leveraging simple and interpretable methods, to improve reasoning capabilities through post-training.
I am also interested in Agentic AI, with a particular focus on empowering local language models to solve complex long-horizon tasks, such as coding and web search, at low cost. My research interests include developing effective agent harnesses with principled memory systems, designing routing strategies that optimize the trade-off between model invocation cost and task performance, and improving communication and collaboration among multiple models. Ultimately, I aim to build efficient and capable agent systems that enable local models to perform complex real-world tasks with minimal reliance on large cloud models.
I am always open to research collaborations. If you are interested in my previous work or would like to discuss potential ideas, please feel free to contact me.
Selected milestones, new work, and announcements.
One paper was accepted by COLM 2026.
One paper was accepted by IEEE TCSVT.
One paper was accepted by ECCV 2026.
I started my internship at NVIDIA, welcome to catch up with me!
Released MemEye: A Visual-Centric Evaluation Framework for Multimodal Agent Memory ↗.
Honored as a Top Reviewer for ICML 2026.
One first-author paper was accepted by ICML 2026.
Joining NVIDIA as a research intern in Summer 2026.
Selected for the Rutgers Climate and Energy Institute Fellowship.
Joined Rutgers University as a Ph.D. student, advised by Prof. Ruixiang Tang.
From Commands to Prompts was accepted by ICLR 2025.
Research across agentic AI, model reasoning, and computer vision.
AGENTIC AI · 2026
MemEye is a vision-centric long-term memory benchmark that evaluates agents’ ability to remember, update, and reason over visual information across long-running, multi-session, image-grounded interactions.
ICML 2026
We investigate the actionable mechanistic interpretation of massive values in large language models, identify the key role of a single layer in their emergence, and introduce a method for mitigating massive activations.
LLM REASONING · 2025
We study how semantically meaningless tokens can produce meaningful reasoning gains through activation shifts, revealing a simple and interpretable mechanism for improving the reasoning capabilities of large language models.
* denotes equal contribution.
RESEARCH AREA
ARXIV
Zeru Shi*, Minghao Guo*, Qingyue Jiao*, Yihao Quan, Boxuan Zhang, Danrui Li, et al.
ARXIV
Boxuan Zhang*, Jianing Zhu*, Zeru Shi, Dongfang Liu, Ruixiang Tang
ICLR 2025
Zeru Shi*, Kai Mei*, Mingyu Jin, Yongye Su, Chaoji Zuo, Wenyue Hua, et al.
EMNLP 2025 · MAIN
Yongye Su, Yucheng Zhang, Zeru Shi, Bruno Ribeiro, Elisa Bertino
RESEARCH AREA
ICML 2026
Zeru Shi, Zhenting Wang, Fan Yang, Qifan Wang, Ruixiang Tang
ARXIV
Zeru Shi*, Kai Mei*, Yihao Quan, Dimitris N. Metaxas, Ruixiang Tang
ARXIV
Zeru Shi, Yingjia Wan, Zhenting Wang, Qifan Wang, Fan Yang, Elisa Kreiss, Ruixiang Tang
ECCV 2026
Liwei Che, Zhiyu Xue, Yihao Quan, Benlin Liu, Zeru Shi, Michelle Hurst, Jacob Feldman, Ruixiang Tang, Ranjay Krishna, Vladimir Pavlovic
COLM 2026
Yihao Quan, Zeru Shi, Jinman Zhao, Ruixiang Tang
RESEARCH AREA
IEEE TCSVT (IF=11.1)
Zeru Shi, Zengxi Zhang, Kemeng Cui, Ruizhe An, Jinyuan Liu, Zhiying Jiang
IEEE TCSVT (IF=11.1)
Zengxi Zhang, Zeru Shi, Jinyuan Liu, Zhiying Jiang
Ph.D. in Computer Science · New Brunswick · NJ · USA
B.E. in Software Engineering · Liaoning · China
Research Intern · Santa Clara · California · USA
Research Intern · Shanghai, China
ICML 2026 Silver Reviewer Award
Rutgers Climate and Energy Institute (RCEI) Fellowship, USA
Dalian Excellent Undergraduate Graduate, China
National Scholarship, China
Special Scholarship of NOK Corporation, Japan
Conference Reviewer: ICLR 2026 · ICML 2026 · NeurIPS 2026 · ACL 2026 · EMNLP 2026 · AAAI 2027 · AC of AI4Math@ICML2026
Journal Reviewer: TOIS