Generative Vision Atlas

Open the atlas · Code

A living literature review of generative vision, organized around one question: where should generation happen — in a VAE latent, a foundation model’s features, discrete tokens, or raw pixels?

• 204 papers in 35 research lines, each defined by the bet it makes
• Benchmark numbers always shown with the settings that produced them — no leaderboard
• Medical imaging covered as a separate domain
• Updated every three days from arXiv; every entry verified against its source