Do Machines Fail Like Humans? A Human-Centred Out-of-Distribution Spectrum for Mapping Error Alignment

Researchers propose a human-centered framework to assess how AI systems align with human decision-making processes by measuring performance on distorted stimuli based on human perceptual difficulty. This approach reveals that vision-language models are most consistently aligned with humans across different levels of challenge, while CNNs and ViTs show varying degrees of alignment depending on the level of distortion.

Read the full article at arXiv cs.AI (Artificial Intelligence)

---

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