LaTtE-Flow: Layerwise Timestep-Expert Flow-based Transformer

• URL: http://arxiv.org/abs/2506.06952v1
• Date: Sun, 08 Jun 2025 00:15:32 GMT
• Title: LaTtE-Flow: Layerwise Timestep-Expert Flow-based Transformer
• Authors: Ying Shen, Zhiyang Xu, Jiuhai Chen, Shizhe Diao, Jiaxin Zhang, Yuguang Yao, Joy Rimchala, Ismini Lourentzou, Lifu Huang,
• Abstract summary: We propose Layerwise Timestep-Expert Flow-based Transformer (LaTtE-Flow), a novel and efficient architecture that unifies image understanding and generation within a single multimodal model.<n>LaTtE-Flow builds upon powerful pretrained Vision-Language Models (VLMs) to inherit strong multimodal understanding capabilities.<n>Experiments demonstrate that LaTtE-Flow achieves strong performance on multimodal understanding tasks, while achieving competitive image generation quality with around 6x faster inference speed.
• Score: 36.51630912419451
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Recent advances in multimodal foundation models unifying image understanding and generation have opened exciting avenues for tackling a wide range of vision-language tasks within a single framework. Despite progress, existing unified models typically require extensive pretraining and struggle to achieve the same level of performance compared to models dedicated to each task. Additionally, many of these models suffer from slow image generation speeds, limiting their practical deployment in real-time or resource-constrained settings. In this work, we propose Layerwise Timestep-Expert Flow-based Transformer (LaTtE-Flow), a novel and efficient architecture that unifies image understanding and generation within a single multimodal model. LaTtE-Flow builds upon powerful pretrained Vision-Language Models (VLMs) to inherit strong multimodal understanding capabilities, and extends them with a novel Layerwise Timestep Experts flow-based architecture for efficient image generation. LaTtE-Flow distributes the flow-matching process across specialized groups of Transformer layers, each responsible for a distinct subset of timesteps. This design significantly improves sampling efficiency by activating only a small subset of layers at each sampling timestep. To further enhance performance, we propose a Timestep-Conditioned Residual Attention mechanism for efficient information reuse across layers. Experiments demonstrate that LaTtE-Flow achieves strong performance on multimodal understanding tasks, while achieving competitive image generation quality with around 6x faster inference speed compared to recent unified multimodal models.

Related papers

• MENTOR: Efficient Multimodal-Conditioned Tuning for Autoregressive Vision Generation Models [30.494968865008513]
  Recent text-to-image models struggle with precise visual control, balancing multimodal inputs, and requiring extensive training for complex image generation.<n>We propose MENTOR, a novel framework for efficient Multimodal-conditioned Tuning for Autoregressive multimodal image generation.<n>Our method delivers superior image reconstruction fidelity, broad task adaptability, and improved training efficiency compared to diffusion-based methods.
  arXiv  Detail & Related papers  (2025-07-13T10:52:59Z)
• Improving Progressive Generation with Decomposable Flow Matching [50.63174319509629]
  Decomposable Flow Matching (DFM) is a simple and effective framework for the progressive generation of visual media.<n>On Imagenet-1k 512px, DFM achieves 35.2% improvements in FDD scores over the base architecture and 26.4% over the best-performing baseline.
  arXiv  Detail & Related papers  (2025-06-24T17:58:02Z)
• Policy Optimized Text-to-Image Pipeline Design [72.87655664038617]
  We introduce a novel reinforcement learning-based framework for text-to-image generation.<n>Our approach first trains an ensemble of reward models capable of predicting image quality scores directly from prompt-workflow combinations.<n>We then implement a two-phase training strategy: initial vocabulary training followed by GRPO-based optimization.
  arXiv  Detail & Related papers  (2025-05-27T17:50:47Z)
• FUDOKI: Discrete Flow-based Unified Understanding and Generation via Kinetic-Optimal Velocities [76.46448367752944]
  multimodal large language models (MLLMs) unify visual understanding and image generation within a single framework.<n>Most existing MLLMs rely on autore (AR) architectures, which impose inherent limitations on future development.<n>We introduce FUDOKI, a unified multimodal model purely based on discrete flow matching.
  arXiv  Detail & Related papers  (2025-05-26T15:46:53Z)
• Deeply Supervised Flow-Based Generative Models [16.953166973699577]
  DeepFlow is a novel framework that enhances velocity representation through inter layer communication.<n>Powered by the improved deep supervision via the internal velocity alignment, DeepFlow converges 8 times faster on ImageNet with equivalent performance.<n>DeepFlow also outperforms baselines in text to image generation tasks, as evidenced by evaluations on MSCOCO and zero shot GenEval.
  arXiv  Detail & Related papers  (2025-03-18T17:58:08Z)
• Upcycling Text-to-Image Diffusion Models for Multi-Task Capabilities [8.81828807024982]
  Multi-Task Upcycling (MTU) is a recipe that extends the capabilities of a pre-trained text-to-image diffusion model to support a variety of image-to-image generation tasks.<n>MTU is the first multi-task diffusion modeling approach that seamlessly blends multi-tasking with on-device compatibility.<n>We show that the performance of MTU is on par with the single-task fine-tuned diffusion models across several tasks.
  arXiv  Detail & Related papers  (2025-03-14T22:19:20Z)
• EMMA: Empowering Multi-modal Mamba with Structural and Hierarchical Alignment [39.870809905905325]
  We propose Empowering Multi-modal Mamba with Structural and Hierarchical Alignment (EMMA) to extract fine-grained visual information. Our model shows lower latency than other Mamba-based MLLMs and is nearly four times faster than transformer-based MLLMs of similar scale during inference.
  arXiv  Detail & Related papers  (2024-10-08T11:41:55Z)
• FlowTurbo: Towards Real-time Flow-Based Image Generation with Velocity Refiner [70.90505084288057]
  Flow-based models tend to produce a straighter sampling trajectory during the sampling process. We introduce several techniques including a pseudo corrector and sample-aware compilation to further reduce inference time. FlowTurbo reaches an FID of 2.12 on ImageNet with 100 (ms / img) and FID of 3.93 with 38 (ms / img)
  arXiv  Detail & Related papers  (2024-09-26T17:59:51Z)
• Versatile Diffusion: Text, Images and Variations All in One Diffusion Model [76.89932822375208]
  Versatile Diffusion handles multiple flows of text-to-image, image-to-text, and variations in one unified model. Our code and models are open-sourced at https://github.com/SHI-Labs/Versatile-Diffusion.
  arXiv  Detail & Related papers  (2022-11-15T17:44:05Z)
• Normalizing Flows with Multi-Scale Autoregressive Priors [131.895570212956]
  We introduce channel-wise dependencies in their latent space through multi-scale autoregressive priors (mAR) Our mAR prior for models with split coupling flow layers (mAR-SCF) can better capture dependencies in complex multimodal data. We show that mAR-SCF allows for improved image generation quality, with gains in FID and Inception scores compared to state-of-the-art flow-based models.
  arXiv  Detail & Related papers  (2020-04-08T09:07:11Z)
• Concurrently Extrapolating and Interpolating Networks for Continuous Model Generation [34.72650269503811]
  We propose a simple yet effective model generation strategy to form a sequence of models that only requires a set of specific-effect label images. We show that the proposed method is capable of producing a series of continuous models and achieves better performance than that of several state-of-the-art methods for image smoothing.
  arXiv  Detail & Related papers  (2020-01-12T04:44:44Z)

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.