The Invisible Gorilla Effect in Out-of-distribution Detection

Researchers identified a new bias in out-of-distribution (OOD) detection called the Invisible Gorilla Effect, where OOD detection performance improves when artefacts share visual similarity with the model's region of interest and declines otherwise. This finding underscores the need for content creators to consider visual similarity between test data and training data regions to enhance the reliability of their models.

Read the full article at arXiv cs.LG (ML)

---

Want to create content about this topic? Use Nemati AI tools to generate articles, social posts, and more.