SFU Trustworthy AI Lab (TAI Lab)
University lab whose stated mission is to "Promote fully Trustworthy and Aligned Artificial General Intelligence, with a focus on certifiability", through theoretically principled approaches. Two named research thrusts: "Principled Understanding" (how capabilities "are correlated and jointly acquired and stored in large deep learning models") and "Certifiable Mitigations" (certification methods combined with data and model refinement). Representative work is certified robustness ("SoK: Certified Robustness for Deep Neural Networks", IEEE S&P 2023; "Transformation-specific Smoothing for Robustness Certification", ACM CCS 2021) and LLM benchmarking ("InfiBench: Benchmarking Question-Answering for Code LLMs", NeurIPS 2024 Datasets and Benchmarks Track).