arXiv
AI/ML
How AI Learns to Spot the Weird Stuff Hidden in Blurry Photos
Imagine a radiologist who can't see the full X-ray, only scattered pixels—but can still flag when something looks off. That's what this method does with diffusion models: it detects when data looks wrong, even in partial measurements.
This means AI systems used in medical imaging or scientific measurement can now catch subtle problems (like equipment drift or patient anomalies) without needing examples of what "wrong" looks like beforehand.
Bug reported: No