Scholar
Junhong Shen
Google Scholar ID: M561o6QAAAAJ
Ph.D. student in Machine Learning, Carnegie Mellon University
Citations & Impact
Citations
512
H-index
12
i10-index
12
Publications
18
Co-authors
14
list available
Publications
Resume (English only)
Academic Achievements
- Paper 'Thinking vs. Doing: Agents that Reason by Scaling Test-Time Interaction' accepted at NeurIPS 2025; paper 'CAT: Content-Adaptive Image Tokenization' accepted at NeurIPS 2025; paper 'Mixture-of-Mamba: Enhancing Multi-Modal State-Space Models with Modality-Aware Sparsity' presented orally at ICLR 2025 Scalable Optimization for Efficient and Adaptive Foundation Models Workshop; Tag-LLM work accepted at ICML 2024.
Research Experience
- Interned at DeepMind working on a code generation agent for visual reasoning; interned at FAIR, contributing to Content-Adaptive Tokenizer (CAT) and Multi-Modal Mixture-of-Mamba projects; developed ScribeAgent web agents based on open-source LLMs.
Education
- Ph.D. student in the Machine Learning Department at Carnegie Mellon University, advised by Ameet Talwalkar; B.S. in Mathematics of Computation from UCLA, where he worked with Lin Yang on sample-efficient reinforcement learning, and also worked on multi-agent RL and Theory of Mind, advised by Song-Chun Zhu and Ying Nian Wu.
Background
- Research interests include enhancing LLMs' interaction with real-world applications, particularly building multi-modal models and agent systems that operate in real-world environments such as browsers, command lines, and IDEs. Also interested in enhancing LLMs' abilities to model diverse data types and applying them to long-tail, low-resource domains like science and business.
Miscellany
- Seeking research scientist positions starting October 2025; supported by J.P. Morgan AI PhD Fellowship; organized the CMU Agent Workshop; participated in organizing the 2022 AutoML Decathlon.
