Normalizing Trajectory Models achieve exact likelihood training with few steps

Researchers have introduced Normalizing Trajectory Models (NTM), a novel approach to generative modeling that maintains exact likelihood training even with a reduced number of sampling steps. NTM integrates conditional normalizing flows within each step of the generative process, allowing for expressive modeling and end-to-end training. This method achieves competitive or superior performance on text-to-image benchmarks using only four sampling steps, a significant improvement in efficiency while preserving the integrity of the likelihood framework. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Introduces a method for more efficient generative modeling with exact likelihood, potentially speeding up image generation and improving model training.

RANK_REASON Publication of a new academic paper detailing a novel modeling approach. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• Normalizing Trajectory Models
• arXiv

• paper
• model release