Best of Both Worlds: Multimodal Reasoning and Generation via Unified Discrete Flow Matching
This addresses multimodal AI challenges for applications requiring joint reasoning and generation, though it appears incremental as it builds on existing flow-matching and adapter techniques.
The paper tackles the problem of multimodal understanding, generation, and editing by proposing UniDFlow, a unified discrete flow-matching framework that decouples tasks via adapters and uses reference-based alignment for improved faithfulness and controllability. It achieves state-of-the-art performance across eight benchmarks and demonstrates strong zero-shot generalization to various tasks without explicit training.
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.