Open and Inclusive
NTUAI welcomes students interested in AI, machine learning, and deep learning regardless of prior experience. We focus on building a supportive space for learning, collaboration, and practical growth.
Diverse Events
We host educational workshops, welcome and mid-term gatherings, and industry trend keynotes to connect members with both technical skills and real-world perspectives.
Project-oriented Learning
NTUAI invites leading researchers and entrepreneurs to help members develop AI applications from zero to hero through hands-on projects, peer sharing, and production-minded experimentation.
Research Highlights
Our members contribute to practical AI research, including benchmark and decoding work such as Multi-TW and Token Constraint Decoding, with a focus on trustworthy and useful AI systems.
Research
Explore recent research initiatives, publications, and benchmark contributions from NTUAI members.
Token Constraint Decoding Improves Robustness on Question Answering for Large Language Models
An inference-time decoding strategy that boosts LLM QA robustness under noisy and adversarial prompt conditions.
Multi-TW — Benchmarking Multimodal Models on Traditional Chinese QA in Taiwan
A Taiwan-centric tri-modal benchmark for Traditional Chinese QA that measures both model accuracy and latency trade-offs.