Welcome! I'm Junjie Chen
Ph.D. in Computer Science. at Tokyo
LLM/NLP applied scientist with a Computer Science Ph.D. from UTokyo and postdoctoral research experience at the University of Oxford. Built LLM-centric multimodal translation and evaluation systems, including AutoMQM with 0.87 correlation to human MQM judgments. Developed a noise-robust audio-based heartbeat detection algorithm, achieving 0.90 r and 0.87 R² on heart rate and RMSSD for high-confidence predictions. Experience in GPU cluster management and supported evaluation and training pipelines for LLMs from 180M to 175B parameters.
For my past experience, please refer to my Resume. Feel free to get in touch via junjiechen.chris@outlook.com
Recent Publications
Improving Unsupervised Constituency Parsing via Maximizing Semantic Information
ICLR • Apr 2025
ProsodyFM: Unsupervised Phrasing and Intonation Control for Intelligible Speech Synthesis
AAAI • Feb 2025
Constituents are Frequent Word Sequences among Sentences with Equivalent Predicate-Argument Structures: Unsupervised Constituency parsing by Span Matching
Findings of ACL • Aug 2024