MMCORE: MultiModal COnnection with Representation Aligned Latent Embeddings

TL;DR

MMCORE is a unified multimodal image generation framework that uses a pre-trained Vision-Language Model to guide diffusion-based visual synthesis, enabling efficient, high-fidelity, and complex multimodal tasks.

Contribution

It introduces a novel approach that transfers VLM understanding into diffusion models without deep fusion or training from scratch, reducing computational costs.

Findings

Outperforms state-of-the-art methods on various benchmarks.

Effectively handles complex multimodal tasks like spatial reasoning.

Maintains high-fidelity image synthesis with reduced computational overhead.

Abstract

We present MMCORE, a unified framework designed for multimodal image generation and editing. MMCORE leverages a pre-trained Vision-Language Model (VLM) to predict semantic visual embeddings via learnable query tokens, which subsequently serve as conditioning signals for a diffusion model. This streamlined design effectively transfers the rich understanding and reasoning capabilities of VLMs into the visual generation process. By obviating the need for deep fusion between autoregressive and diffusion models or training from scratch, MMCORE significantly reduces computational overhead while maintaining high-fidelity synthesis. MMCORE seamlessly integrates text-to-image synthesis with interleaved image generation, demonstrating robust multimodal comprehension in complex scenarios such as spatial reasoning and visual grounding. Comprehensive evaluations indicate that MMCORE consistently outperforms state-of-the-art baselines across a broad spectrum of text-to-image and single/multi-image editing benchmarks.