Scholar
Lichang Chen
Google Scholar ID: XesgJyUAAAAJ
University of Maryland
Citations & Impact
Citations
2,097
H-index
18
i10-index
22
Publications
20
Co-authors
15
list available
Publications
Resume (English only)
Academic Achievements
- Published papers: 'OmnixR: Evaluating Omni-modality Language Models on Reasoning across Modalities' at ICLR 2025, 'From Lists to Emojis: How Format Bias Affects Model Alignment' at ACL 2025, 'RRM: Robust Reward Model Training Mitigates Reward Hacking' at ICLR 2025, 'ODIN: Disentangled Reward Mitigates Hacking in RLHF' at ICML 2024, 'InstructZero: Efficient Instruction Optimization for Black-Box Large Language Models' at ICML 2024, 'AlpaGasus: Training a Better Alpaca with Fewer Data' at ICLR 2024, 'Backdooring Instruction-Tuned Large Language Models with Virtual Prompt Injection' at NAACL 2024, 'HallusionBench: an image-context reasoning benchmark challenging for multi-modality models' at CVPR 2024, 'How Many Demonstrations Do You Need for In-context Learning?' at EMNLP 2023, 'PTP: Boosting Stability and Performance of Prompt Tuning with Perturbation-Based Regularizer' at EMNLP 2023.
Research Experience
- AI Research Scientist at Meta SuperIntelligence, 2025.8 - Present, working on agents for automating AI researchers; Intern at Google Deepmind, 2024.9 - 2025.5, focused on Meta RL for reasoning & Thinking to Learn; Intern at Google Deepmind, 2024.5 - 2024.8, worked on the evaluation/alignment of Omni-Modality Language Models; Intern at Google Research & Cloud AI Research, 2024.2 - 2024.5, involved in the self-improvement of multimodal LLMs; Intern at NVIDIA ADLR team, 2023.9 - 2024.1, mitigated hackings in RLHF/developed more robust reward models; Intern at Samsung AI Research, 2023.5 - 2023.8, created a data filter for instruction tuning.
Education
- PhD: Computer Science Department, University of Maryland, College Park, Advisor: Dr. Tianyi Zhou; Bachelor's Degree: Zhejiang University.
Background
- Research interests lie in AI alignment and agentic systems, particularly in automating realistic workflows such as foundational model research, machine learning, and data analysis. Also interested in creating scalable RL environments to improve AI capabilities like reasoning and instruction following.
Miscellany
- Contact: {bob}{my-last-name}@cs.umd.edu; Social media links: Google Scholar, Twitter, LinkedIn, Github
