Exploring MLLM-Diffusion Information Transfer with MetaCanvas

• URL: http://arxiv.org/abs/2512.11464v1
• Date: Fri, 12 Dec 2025 11:07:11 GMT
• Title: Exploring MLLM-Diffusion Information Transfer with MetaCanvas
• Authors: Han Lin, Xichen Pan, Ziqi Huang, Ji Hou, Jialiang Wang, Weifeng Chen, Zecheng He, Felix Juefei-Xu, Junzhe Sun, Zhipeng Fan, Ali Thabet, Mohit Bansal, Chu Wang,
• Abstract summary: We propose a lightweight framework that lets MLLMs reason and plan directly in spatial and multimodal latent spaces.<n>We evaluate it across six visual generation tasks, including text-to-image generation, text/image-to-video generation, image/video attribute editing, and in-context video generation.
• Score: 66.28602082523464
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Multimodal learning has rapidly advanced visual understanding, largely via multimodal large language models (MLLMs) that use powerful LLMs as cognitive cores. In visual generation, however, these powerful core models are typically reduced to global text encoders for diffusion models, leaving most of their reasoning and planning ability unused. This creates a gap: current multimodal LLMs can parse complex layouts, attributes, and knowledge-intensive scenes, yet struggle to generate images or videos with equally precise and structured control. We propose MetaCanvas, a lightweight framework that lets MLLMs reason and plan directly in spatial and spatiotemporal latent spaces and interface tightly with diffusion generators. We empirically implement MetaCanvas on three different diffusion backbones and evaluate it across six tasks, including text-to-image generation, text/image-to-video generation, image/video editing, and in-context video generation, each requiring precise layouts, robust attribute binding, and reasoning-intensive control. MetaCanvas consistently outperforms global-conditioning baselines, suggesting that treating MLLMs as latent-space planners is a promising direction for narrowing the gap between multimodal understanding and generation.

Related papers

• Graph-MLLM: Harnessing Multimodal Large Language Models for Multimodal Graph Learning [23.089644598166885]
  Multimodal Large Language Models (MLLMs) have demonstrated remarkable capabilities in representing and understanding diverse modalities.<n>Integrating multimodality with structured graph information (i.e., multimodal graphs, MMGs) is essential for real-world applications such as social networks, healthcare, and recommendation systems.<n>Existing MMG learning methods fall into three paradigms based on how they leverage MLLMs.
  arXiv  Detail & Related papers  (2025-06-12T01:44:46Z)
• Slot-MLLM: Object-Centric Visual Tokenization for Multimodal LLM [21.967692616735196]
  multimodal large language models (MLLMs) have emerged as a key approach in achieving artificial general intelligence.<n>We propose an object-centric visual tokenizer based on Slot Attention specifically for MLLMs.<n>This work is the first demonstration of the feasibility of object-centric slot attention performed with MLLMs and in-the-wild natural images.
  arXiv  Detail & Related papers  (2025-05-23T10:43:45Z)
• Generative Multimodal Pretraining with Discrete Diffusion Timestep Tokens [66.02261367232256]
  Multimodal Large Language Models (MLLMs) aim to unify visual comprehension and generation.<n>Existing approaches rely on spatial visual tokens, where image patches are encoded and arranged according to a spatial order.<n>In this paper, we build a proper visual language by reconstructing diffusion timesteps to learn discrete visual tokens.
  arXiv  Detail & Related papers  (2025-04-20T16:14:28Z)
• Transfer between Modalities with MetaQueries [44.57406292414526]
  We introduce MetaQueries, a set of learnable queries that act as an efficient interface between autoregressive multimodal LLMs and diffusion models.<n>Our method simplifies training, requiring only paired image-caption data and standard diffusion objectives.<n>Our method is flexible and can be easily instruction-tuned for advanced applications such as image editing and subject-driven generation.
  arXiv  Detail & Related papers  (2025-04-08T17:58:47Z)
• ILLUME+: Illuminating Unified MLLM with Dual Visual Tokenization and Diffusion Refinement [68.05833403672274]
  Existing unified models have struggled to handle the three fundamental capabilities in a unified model: understanding, generation, and editing.<n>ILLUME+ introduces a unified dual visual tokenizer, DualViTok, which preserves fine-grained textures and text-aligned semantics.<n>We also employ a diffusion model as the image detokenizer for enhanced generation quality and efficient super-resolution.
  arXiv  Detail & Related papers  (2025-04-02T17:45:00Z)
• Rethinking Visual Prompting for Multimodal Large Language Models with External Knowledge [76.45868419402265]
  multimodal large language models (MLLMs) have made significant strides by training on vast high-quality image-text datasets. However, the inherent difficulty in explicitly conveying fine-grained or spatially dense information in text, such as masks, poses a challenge for MLLMs. This paper proposes a new visual prompt approach to integrate fine-grained external knowledge, gleaned from specialized vision models, into MLLMs.
  arXiv  Detail & Related papers  (2024-07-05T17:43:30Z)
• Multi-modal Instruction Tuned LLMs with Fine-grained Visual Perception [63.03288425612792]
  We propose bfAnyRef, a general MLLM model that can generate pixel-wise object perceptions and natural language descriptions from multi-modality references. Our model achieves state-of-the-art results across multiple benchmarks, including diverse modality referring segmentation and region-level referring expression generation.
  arXiv  Detail & Related papers  (2024-03-05T13:45:46Z)
• Probing Multimodal Large Language Models for Global and Local Semantic Representations [57.25949445963422]
  We study which layers of Multimodal Large Language Models make the most effort to the global image information. In this study, we find that the intermediate layers of models can encode more global semantic information. We find that the topmost layers may excessively focus on local information, leading to a diminished ability to encode global information.
  arXiv  Detail & Related papers  (2024-02-27T08:27:15Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.