Stanford CS 224N

#artificialintelligence 

There were three options for the course final project. Students either chose their own topic as a custom final project, or else they took part in one of the options for the default final projects, involving building question-answering systems. This year, we had two default final project options: Either people could build from scratch (regular, IID) question-answering models for the SQuAD 2.0 challenge or in the Robust QA track, students started with 3 question-answering datasets (SQuAD, Natural Questions, and NewsQA) and a pre-trained, transformer QA system and worked to produce a system that worked robustly on (OOD) test sets from additional domains. You can find links to previous years' reports under Previous Offerings on the homepage.