AI 11
- Numbers Eat Pipelines: Two Projects, One Habit of Distrusting Your Own Metrics
- How to Actually Report FID: A Single-GPU Reproducibility Checklist
- Can DiffAugment Break the FID-160 Ceiling? A Limited-Data Remedy on a 2,490-Image CLIP-GAN
- Killing a Hypothesis Cheaply: The Discriminator Wasn't Our GAN's Bottleneck
- MS-CLIP-GAN Architecture: How a CLIP-Guided Multi-Stage GAN Is Wired
- [Troubleshooting] Your FID of 0.24 Isn't Near-Perfect — It's the Wrong Feature Space
- Making an ESM2 Protein Variant Classifier Practical
- Held-Out in Name Only: Two Leaks and a Confound in My Detector's Evaluation
- When 0.91 mAP Still Fails: Fixing Safety-Gear False Positives with Empty-GT Negatives
- A Classical Skull-Mask Baseline, and What 5 MRI Slices Did Not Prove
- Building a Protein Variant Classifier with ESM2 and Multi-GPU Training