What happened: Normalizing Trajectory Models

Apple researchers introduced Normalizing Trajectory Models, which generate images in four steps while maintaining exact likelihood through conditional normalizing flows.

Why it matters: Normalizing Trajectory Models

Relevant for engineers building fast generative models who need both speed and likelihood guarantees for image generation tasks.

Watch out: Normalizing Trajectory Models

The approach requires architectural changes combining invertible blocks with deep predictors, and real-world performance gains versus existing methods need independent verification.

In the news

Normalizing Trajectory Models

Apple Machine Learning Research · Published October 8, 2026 · 3 min read

In 30 seconds

What happened: Apple researchers introduced Normalizing Trajectory Models, which generate images in four steps while maintaining exact likelihood through conditional normalizing flows.
Why it matters: Relevant for engineers building fast generative models who need both speed and likelihood guarantees for image generation tasks.
Watch out: The approach requires architectural changes combining invertible blocks with deep predictors, and real-world performance gains versus existing methods need independent verification.

Read the full article

distill

The patterns behind this

Each one covers how the technique works, when it earns its cost, and where it breaks.

The Agent Architect

One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.

Weekly email, one-click unsubscribe. We only use your address to send the briefing.