JanusFlow: Harmonizing Autoregression and Rectified Flow for Unified Multimodal Understanding and Generation

TL;DR

JanusFlow is a unified model combining autoregressive language modeling and rectified flow for improved multimodal understanding and generation, achieving high performance without complex architecture changes.

Contribution

It introduces a minimalist architecture that unifies image understanding and generation by integrating autoregressive models with rectified flow, with novel training strategies for decoupling and aligning encoders.

Findings

Achieves comparable or superior performance to specialized models.

Outperforms existing unified approaches on standard benchmarks.

Demonstrates rectified flow can be trained within large language models.

Abstract

We present JanusFlow, a powerful framework that unifies image understanding and generation in a single model. JanusFlow introduces a minimalist architecture that integrates autoregressive language models with rectified flow, a state-of-the-art method in generative modeling. Our key finding demonstrates that rectified flow can be straightforwardly trained within the large language model framework, eliminating the need for complex architectural modifications. To further improve the performance of our unified model, we adopt two key strategies: (i) decoupling the understanding and generation encoders, and (ii) aligning their representations during unified training. Extensive experiments show that JanusFlow achieves comparable or superior performance to specialized models in their respective domains, while significantly outperforming existing unified approaches across standard benchmarks. This work represents a step toward more efficient and versatile vision-language models.