Machine Learning with a Honk

40. Vision Transformers Need Registers

Fighting global information smuggling as models grow smarter

Oct 1 • 

Massimiliano Viola

8

September 2025

39. From DINO to DINOv3

Evolution of self-supervised vision transformers

Sep 24 • 

Massimiliano Viola

17

2

38. Stable Diffusion VAE is Flawed

Serious claims on Reddit, check this out!

Sep 17 • 

Massimiliano Viola

4

5

37. Step1X-Edit

How to build a dataset for text-guided image editing

Sep 10 • 

Massimiliano Viola

1

36. DIFT: DIffusion FeaTures

Extracting semantic and geometric correspondences from diffusion models

Sep 3 • 

Massimiliano Viola

7

August 2025

35. Segment Anything Model (SAM)

Cut out any object, in any image, with just a few clicks

Aug 27 • 

Massimiliano Viola

4

34. LLaVA: Teaching LLMs to See

Visual instruction tuning to create vision-language models

Aug 20 • 

Massimiliano Viola

33. Visual Anagrams

Generating multi-view optical illusions with diffusion models

Aug 13 • 

Massimiliano Viola

2

2

32. I-JEPA

Self-supervised learning on images with a joint-embedding predictive architecture

Aug 6 • 

Massimiliano Viola

July 2025

31. Marigold

Repurposing diffusion-based image generators for dense prediction tasks

Jul 30 • 

Massimiliano Viola

1

30. IP-Adapter

Enable image prompt capabilities for pre-trained text-to-image diffusion models

Jul 23 • 

Massimiliano Viola

29. DreamBooth

Fine-tuning diffusion models for subject-driven generation

Jul 16 • 

Massimiliano Viola