FFJORD: Free-form continuous dynamics for scalable reversible generative models

OpenAI Blog 10/02/18, 07:00 AM Papers

Summary

FFJORD introduces a scalable reversible generative model using continuous dynamics and Hutchinson's trace estimator to enable unbiased log-density estimation without architectural constraints. The method achieves state-of-the-art results on density estimation and image generation while maintaining efficient sampling.

No content available

Original Article

View Cached Full Text

Cached at: 04/20/26, 02:45 PM

# FFJORD: Free-form continuous dynamics for scalable reversible generative models Source: [https://openai.com/index/ffjord/](https://openai.com/index/ffjord/) ## Abstract A promising class of generative models maps points from a simple distribution to a complex distribution through an invertible neural network\. Likelihood\-based training of these models requires restricting their architectures to allow cheap computation of Jacobian determinants\. Alternatively, the Jacobian trace can be used if the transformation is specified by an ordinary differential equation\. In this paper, we use Hutchinson's trace estimator to give a scalable unbiased estimate of the log\-density\. The result is a continuous\-time invertible generative model with unbiased density estimation and one\-pass sampling, while allowing unrestricted neural network architectures\. We demonstrate our approach on high\-dimensional density estimation, image generation, and variational inference, achieving the state\-of\-the\-art among exact likelihood methods with efficient sampling\.

Similar Articles

DiffoR: A Unified Continuous Generative Framework for Universal Ordinal Regression

arXiv cs.LG

DiffoR proposes a novel continuous generative framework for ordinal regression using diffusion models, overcoming limitations of discrete methods. Extensive experiments on 12 benchmarks demonstrate state-of-the-art performance across four domains.

x-Prediction Is All You Need:Training-Free Accelerated Generation via Endpoint Decodability

arXiv cs.LG

This paper introduces Truncated Jump Sampling (TJS), a training-free method that accelerates diffusion and flow matching model generation by exploiting endpoint decodability, reducing neural function evaluations by 20–70% with near-matched quality across multiple models.

FFAvatar: Few-Shot, Feed-Forward, and Generalizable Avatar Reconstruction

Hugging Face Daily Papers

FFAvatar proposes a feed-forward framework for reconstructing high-quality, animatable 3D Gaussian head avatars from few unposed images in seconds, achieving a 5.5 PSNR improvement over state-of-the-art on the NeRSemble benchmark.

Flow-Direct: Feedback-Efficient and Reusable Guidance for Flow Models via Non-Parametric Guidance Field

arXiv cs.LG

Flow-Direct introduces a non-parametric guidance field for flow-based generative models that accumulates reward feedback persistently, improving feedback efficiency and enabling reuse of collected samples to guide generation for multiple objectives without additional reward evaluations.

Renormalization Group Flow Matching for Scalable Local Generative Modeling

arXiv cs.LG

Introduces renormalization group flow matching (RGFM), a generative framework that uses renormalization group flow for scalable local generative modeling, improving global coherence and computational efficiency.

Similar Articles

DiffoR: A Unified Continuous Generative Framework for Universal Ordinal Regression

x-Prediction Is All You Need:Training-Free Accelerated Generation via Endpoint Decodability

FFAvatar: Few-Shot, Feed-Forward, and Generalizable Avatar Reconstruction

Flow-Direct: Feedback-Efficient and Reusable Guidance for Flow Models via Non-Parametric Guidance Field

Renormalization Group Flow Matching for Scalable Local Generative Modeling

Submit Feedback