Scholar
Xinliang Frederick Zhang
Google Scholar ID: -uGCT5QAAAAJ
PhD Candidate, University of Michigan
Citations & Impact
Citations
277
H-index
8
i10-index
7
Publications
16
Co-authors
7
list available
Publications
Resume (English only)
Academic Achievements
- Outstanding Reviewer awards at EMNLP 2024 and NAACL 2022.
- Best Paper Award at BIBM 2021.
- Paper “FINDR: A Fast Influential Data Selector for NL2Code Pretraining” accepted to IJCNLP-AACL 2025.
- Co-authored preprint with Google DeepMind: “Do LLMs Really Need 10+ Thoughts for ‘find the time 1000 days after today?’ Towards Structural Understanding of LLM Overthinking”.
- Preprint: “Logit Arithmetic Elicits Long Reasoning Capabilities Without Training”.
Background
- From Hangzhou, China; Chinese name: 张心亮; commonly goes by Frederick.
- Final-year Ph.D. candidate in Computer Science and Engineering (CSE) at the University of Michigan, advised by Lu Wang; member of LAUNCH Lab and affiliated with Michigan AI Lab.
- Research interests include Natural Language Processing (NLP), Large Language Models (LLM), Deep Learning (DL), Machine Learning (ML), and Computational Social Science (CSS).
- Currently investigating LLM personalization (via memory mechanisms), System-2 thinking (e.g., structured/personalized/over-thinking), and efficient LLMs (training and inference).
- Earlier Ph.D. work focused on structured learning, multi-granularity information extraction, and narrative understanding; undergraduate research included natural language inference, information retrieval, and question answering/generation.
- Also interested in code models, annotation disagreements, and NLP applications in social science and healthcare (clinical NLP).
- Regular reviewer for NLP conferences.
