AI & ML interests

Interpretability for Generative Language Models 🔎 🐛

Inseq 🐛 is a Pytorch-based hackable toolkit to democratize access to common post-hoc interpretability analyses for decoder-only and encoder-decoder sequence generation models.
The Inseq organization on the 🤗 Hub aims to host valuable models, datasets and demos to improve the reproducibility of interpretability studies in the field of natural language generation.

Github Repository
Demo Paper
Documentation