CyCLeGen: Cycle-Consistent Layout Prediction and Image Generation in Vision Foundation Models

TL;DR

CyCLeGen is a unified vision-language model that integrates image understanding and generation through cycle-consistent learning, improving reasoning and data efficiency, and achieving strong results across multiple benchmarks.

Contribution

It introduces a fully integrated autoregressive model with cycle consistency for joint image understanding and generation, enabling self-reflection and data-efficient learning.

Findings

Significant performance improvements on diverse benchmarks

Enhanced reasoning capabilities through cycle consistency

Effective self-supervised learning via synthetic supervision

Abstract

We present CyCLeGen, a unified vision-language foundation model capable of both image understanding and image generation within a single autoregressive framework. Unlike existing vision models that depend on separate modules for perception and synthesis, CyCLeGen adopts a fully integrated architecture that enforces cycle-consistent learning through image->layout->image and layout->image->layout generation loops. This unified formulation introduces two key advantages: introspection, enabling the model to reason about its own generations, and data efficiency, allowing self-improvement via synthetic supervision under a reinforcement learning objective guided by cycle consistency. Extensive experiments show that CyCLeGen achieves significant gains across diverse image understanding and generation benchmarks, highlighting the potential of unified vision-language foundation models.