Biography
I am an MRes student in Electrical Engineering at the University of Nottingham Ningbo China. I am also an Agent Application R&D Intern at LuckyShort, where I work on video-algorithm applications and agent workflows for AI-native content production.
I work on trustworthy artificial intelligence. My research interests broadly lie in multimodal large language models, numerical and time-series reasoning, reliable forecasting, and generative media systems. I am particularly interested in models that can recognize uncertainty, explain their reasoning, and remain dependable when real-world data is noisy or unfamiliar.
Previously, I was an Algorithm R&D Intern at the International Digital Economy Academy (IDEA) and a Research Assistant at the Hong Kong University of Science and Technology (Guangzhou). My work has appeared at venues including EMNLP Findings, ICML, ACM Multimedia, ICRA, ICASSP, IEEE Sensors Journal, and Pattern Recognition.
I am open to research collaborations and engineering discussions on trustworthy AI, multimodal reasoning, time-series intelligence, and agentic content systems.
Recent News
- Our paper NUM2EVENT was accepted to EMNLP 2026 Findings.
- Joined LuckyShort as an Agent Application R&D Intern.
- Our work on reliable time-series forecasting under future uncertainty was accepted to ICASSP 2026.
- Our paper on faithful multimedia web forecasting was published at ACM Multimedia 2025.
- Joined IDEA as an Algorithm R&D Intern, focusing on multimodal LLM post-training.
- Co-authored papers were published at ICML 2025 and ICRA 2025.
- Joined HKUST(GZ) as a Research Assistant.
Selected Publications
Full list on Google Scholar[EMNLP Findings '26] NUM2EVENT: Structured Event Inference from Numerical Time-Series Paper
Highlight: A number-to-event framework that infers structured, interpretable events directly from numerical signals.
[ICASSP '26] Towards Reliable Time Series Forecasting under Future Uncertainty: Ambiguity and Novelty Rejection Mechanisms Paper
Highlight: Reliable forecasting that can identify ambiguous predictions and reject unfamiliar future conditions.
[ACM MM '25] From Guesswork to Guarantee: Towards Faithful Multimedia Web Forecasting with TimeSieve Project
Highlight: Faithfulness-aware optimization that improves robustness to random seeds and several kinds of model perturbation.
[ICML '25] IMTS is Worth Time × Channel Patches: Visual Masked Autoencoders for Irregular Multivariate Time Series Prediction Paper
Highlight: Visual masked autoencoders that model irregular multivariate time series through time-by-channel patches.
[IEEE Sensors Journal] Determination of Tea Quality: A Separate Domain Graph Convolution Network Combined With an Electronic Nose Paper
Highlight: Graph-based feature learning for tea-quality recognition from electronic-nose sensor data.
[Preprint '24] TimeSieve: Extracting Temporal Dynamics through Information Bottlenecks Paper
Highlight: Wavelet decomposition and information bottlenecks for learning useful multi-scale temporal patterns.
Experience
Agent Application R&D Intern, LuckyShort
Video generation, agent workflows, production monitoring, and AI short-drama tooling.
Algorithm R&D Intern, IDEA
Multimodal LLM post-training and structured reasoning from numerical time series.
Research Assistant, HKUST(GZ)
Trustworthy AI, time-series forecasting, and non-intrusive load monitoring.
Undergraduate Researcher, Bionic Sensing & Pattern Recognition Research Team
Graph neural networks and machine olfaction.
Selected engineering impact at LuckyShort
I contributed to script parsing, character assets, video generation, super-resolution, background music, monitoring, and delivery tooling. The median script-processing time decreased from 232 seconds to 145 seconds: a typical task became 87 seconds faster, which is a 37.5% reduction. The BGM failure rate decreased by 35% relative to the previous pipeline, and the rollout data pipeline captured 37,000 real generate and regenerate interactions for analysis.
Education & Awards
MRes in Electrical Engineering
University of Nottingham Ningbo China
B.Eng. in Measurement & Control Technology and Instruments
Northeast Electric Power University
- National College Students' Innovation and Entrepreneurship Training Program project for a real-time meat-quality analysis system based on machine olfaction.
- China Southern Power Grid 2024 Smart Metering Ecosystem Development Competition.
Snapshots of Life
Outside work, skiing is one of my favorite ways to spend winter. During my undergraduate years at Northeast Electric Power University, Tianding Mountain Ski Resort was only about an hour away.
At the beginning of 2026, I travelled with classmates to Haikou and Sanya, Hainan. The trip gave us time to enjoy the coast and start the year together.
