Zheda Mai

Zheda Mai works in computer science on machine learning topics including how to adapt pretrained vision and language models to new tasks efficiently, continual and few-shot learning, and multimodal systems that combine vision, language, and time-series data. Recent work covers areas such as prompt-based model tuning, weakly-supervised image segmentation, embodied navigation agents, and biases in AI-driven recommendation systems. This research is largely methods-focused, aiming to make large pretrained models more adaptable, efficient, and reliable for downstream applications.

Publication output was minimal before 2020, rose to a steady 5-7 papers per year from 2020-2023, dipped in 2024, then increased sharply in 2025-2026, suggesting recent growth in activity.

Generated by claude-sonnet-5 from public bibliographic data · Jul 20, 2026