Interpretability for sequence generation models 🐛 🔍

Explainable AI in Julia.

Code for our ICML '19 paper: Neural Network Attributions: A Causal Perspective.

On Explaining Your Explanations of BERT: An Empirical Study with Sequence Classification

Easy-to-use MIRAGE code for faithful answer attribution in RAG applications. Paper: https://aclanthology.org/2024.emnlp-main.347/

surrogate quantitative interpretability for deepnets

Attribution (or visual explanation) methods for understanding video classification networks. Demo codes for WACV2021 paper: Towards Visually Explaining Video Understanding Networks with Perturbation.

Code for the paper: Towards Better Understanding Attribution Methods. CVPR 2022.

Metrics for evaluating interpretability methods.

The source code for the journal paper: Spatio-Temporal Perturbations for Video Attribution, TCSVT-2021

Source code for the GAtt method in "Faithful and Accurate Self-Attention Attribution for Message Passing Neural Networks via the Computation Tree Viewpoint".

Hacking SetFit so that it works with integrated gradients.

squid repository for manuscript analysis

Code for our AISTATS '22 paper: Improving Attribution Methods by Learning Submodular Functions.

Code for my Master Thesis titled "Exploring Data Augmentation Methods through Attribution".