Best of Both Worlds: Multimodal Reasoning and Generation via Unified Discrete Flow Matching

TL;DR

UniDFlow introduces a unified framework for multimodal understanding and generation that enhances performance, controllability, and generalization across diverse tasks without extensive retraining.

Contribution

It presents a novel discrete flow-matching approach with task-specific adapters and reference-based alignment, enabling versatile multimodal capabilities and state-of-the-art results.

Findings

Achieves SOTA on eight benchmarks

Exhibits strong zero-shot generalization

Improves faithfulness and controllability

Abstract

We propose UniDFlow, a unified discrete flow-matching framework for multimodal understanding, generation, and editing. It decouples understanding and generation via task-specific low-rank adapters, avoiding objective interference and representation entanglement, while a novel reference-based multimodal preference alignment optimizes relative outcomes under identical conditioning, improving faithfulness and controllability without large-scale retraining. UniDFlpw achieves SOTA performance across eight benchmarks and exhibits strong zero-shot generalization to tasks including inpainting, in-context image generation, reference-based editing, and compositional generation, despite no explicit task-specific training.