Goto

Collaborating Authors

 disagreement notion


Beyond Top-Class Agreement: Using Divergences to Forecast Performance under Distribution Shift

Schirmer, Mona, Zhang, Dan, Nalisnick, Eric

arXiv.org Artificial Intelligence

Knowing if a model will generalize to data 'in the wild' is crucial for safe deployment. To this end, we study model disagreement notions that consider the full predictive distribution - specifically disagreement based on Hellinger distance, Jensen-Shannon and Kullback-Leibler divergence. We find that divergence-based scores provide better test error estimates and detection rates on out-of-distribution data compared to their top-1 counterparts. Experiments involve standard vision and foundation models.

  Country:
  Genre: Research Report (0.64)