OneFlow: Concurrent Mixed-Modal and Interleaved Generation with Edit Flows
This addresses the problem of inefficient and sequential multimodal generation for AI applications, representing a novel paradigm rather than an incremental improvement.
The authors tackled the problem of rigid causal ordering in multimodal generation by developing OneFlow, a non-autoregressive model that enables concurrent mixed-modal text-image synthesis. They demonstrated that OneFlow outperforms autoregressive baselines on generation and understanding tasks while using up to 50% fewer training FLOPs across model sizes from 1B to 8B.
We present OneFlow, the first non-autoregressive multimodal model that enables variable-length and concurrent mixed-modal generation. Unlike autoregressive models that enforce rigid causal ordering between text and image generation, OneFlow combines an insertion-based Edit Flow for discrete text tokens with Flow Matching for image latents. OneFlow enables concurrent text-image synthesis with hierarchical sampling that prioritizes content over grammar. Through controlled experiments across model sizes from 1B to 8B, we demonstrate that OneFlow outperforms autoregressive baselines on both generation and understanding tasks while using up to 50% fewer training FLOPs. OneFlow surpasses both autoregressive and diffusion-based approaches while unlocking new capabilities for concurrent generation, iterative refinement, and natural reasoning-like generation.