Shap’s partition explainer for language models

The Shapley value, the Owen value, and the partition explainer in shap: how it all relates

The ability to understand a model’s prediction is often crucial to pave its way into production. While simple, interpretable models achieve good enough results in some applications, the benefit of using complex modeling techniques outweighs the quest for tractability in other applications, like natural language processing or computer vision. Yet, we want to understand which features are most important …

editors pick explainer language language models shap