Glossary term
Glossary term
Evaluation and Benchmarks
Abbreviation for the Learning Interpretability Tool (LIT), which was previously known as the Language Interpretability Tool.
Created for this library
A research team uses LIT to inspect attention and saliency maps on misclassified examples during model review.
An ML platform team uses LIT in release walkthroughs so reviewers can explore prediction explanations interactively.
An NLP team uses LIT to compare two model variants on the same prompts side by side during prerelease evaluation.
Definition source: Google for Developers Machine Learning Glossary | Creative Commons Attribution 4.0 License