Programming Assignments

We plan to have 4 programming homework assignments, covering the first part of the course (before the fall break). The first homework focuses on training a model, while the remaining assignments require using explainability methods we will talk about in the lectures to explain the predictions of your model. The goal of these assignments is to prepare you for a successful project that goes beyond trying out the method which will be the focus of the second part of the course.

1. Assignment #1:
   • Model finetuning
2. Assignment #2:
   • Prompting
   • Free-text explanations
   • Model uncertainty
3. Assignment #3:
   • Gradient-based input attribution
   • Attention-based explanations
   • Concept-based explanations
4. Assignment #4:
   • Contrastive editing
   • Data influence

Assignment instructions will be shared on Canvas.