arXiv Computer Vision

RepFusion: Leveraging Multimodal Priors for Denoising in Representation Space

RepFusion leverages multimodal large language models (MLLMs) as noisy representation encoders to condition a diffusion transformer for denoising in representation space. By reusing MLLM outputs across denoising steps, the method achieves faster inference while maintaining performance comparable to per-step recomputation. Experiments show RepFusion outperforms baselines with similar inference budgets, demonstrating that MLLMs provide strong priors for denoising visual representations.

arXiv Machine Learning
Sep 24

Repurposing Pre-trained LLMs as High Fidelity Continuous Text Autoencoders

The paper introduces LLMAE, a technique that transforms a pretrained decoder-only language model into a continuous text autoencoder by inserting a fixed-length latent bottleneck into its internal activations. Using a 270M Gemma 3 model with structured attention masks, LoRA adaptation, and KL regularization, LLMAE achieves near-perfect reconstruction of text sequences up to 1024 tokens. The authors further show that the resulting latent representation can be leveraged to train a latent text diffusion model for detailed image captioning, demonstrating downstream utility.

By Arkanath Pathak, Unnat Jain, Alexander C. Berg
arXiv AI
3d ago

DRAMA: Dissecting Attention Redundancy for Accelerating Multimodal Diffusion Large Language Models

DRAMA is a training‑free acceleration framework for multimodal diffusion large language models that dissects attention redundancy by probing attention tensors along spatial and temporal axes. It identifies two key redundancy patterns: a sparse subset of visual tokens consistently receives most attention, and attention distributions stabilize quickly after an initial phase. Leveraging these insights, DRAMA introduces Spatial‑Redundancy‑Aware Visual Token Pruning and Sufficiency‑Guided Adaptive Stopping, achieving an average 18.67× inference speedup across six vision‑language benchmarks while preserving competitive accuracy.

By Yi-Hsin Hung, Fangfu Liu, Shao-Yuan Lo, Chu-Song Chen
arXiv Computer Vision
Oct 12

InstantFusion: A Shared Latent Interface for Heterogeneous Image Generators

InstantFusion introduces a lightweight shared latent interface that enables cross-model state translation between heterogeneous image generators. By training model-specific Latent AutoEncoders to project intermediate denoising states into a common latent space, the method aligns states from the same image at matched noise levels and synchronizes their local denoising velocities. This allows seamless denoising transitions across frozen generators, supporting applications such as acceleration, multi-preference composition, on-policy distillation, and cross-model composition.

By Xuanlang Dai, Jiazi Bu, Yujie Zhou, Yuhong Liu, Beichen Zhang, Penghui Yang, Ziyu Liu, Yuhang Zang
arXiv Computation and Language
Aug 28

Visual Information-Guided Parallel Decoding for Diffusion Multimodal Large Language Models

Visual Information-Guided Parallel Decoding for Diffusion Multimodal Large Language Models introduces the VIG‑Sampler, a method that prioritizes tokens for decoding based on their attention to image tokens and penalizes redundancy in image‑attention distributions. The approach aims to improve the quality of multimodal generation by selecting more informative tokens during diffusion decoding. Experiments on seven captioning and VQA benchmarks with three open‑source dMLLMs show that VIG‑Sampler outperforms the Info‑Gain Sampler by an average of 19.3 CIDEr points and achieves better COCO Caption results using only half as many decoding steps.

By Insu Lee, Wooje Park, Wonseok Shin, Jinwoo Son, Byonghyo Shim
arXiv Machine Learning
Jul 2

Information-Regularized Attention for Visual-Centric Reasoning

arXiv:2607. 00434v1 Announce Type: cross Abstract: Vision-language models (VLMs) have become a paradigm for multimodal learning, yet remain unstable due to object hallucination, weak visual grounding, and catastrophic forgetting after full-parameter instruction tuning.

By Guohao Sun, Xiaofang Wang, Yash Patel, Mengchen Liu, Zhiqiang Tao, Praveen Krishnan
arXiv Machine Learning
Oct 8

ORCA: Hunting Compositional Failures in Text-to-Image Diffusion

ORCA (Orthogonal Residual Compositional Alignment) is a method that improves text-to-image diffusion models by aligning the diffusion transformer’s latent space with a low‑rank target derived from a frozen visual encoder. It introduces an auxiliary loss that uses a predictor with an orthogonal basis parameterised by a learned residual between T5 and CLIP embeddings, providing a prompt‑dependent signal for selecting the visual readout subspace. Experiments on three diffusion‑transformer backbones show that ORCA improves FID and GenEval scores, especially on attribute binding, spatial relations, and multi‑object prompts, without adding inference‑time cost.

By Arshia Hemmat, Amirhossein Vahidi, Amitis Shidani, Mohammad Vali Sanian, Hesam Asadollahzadeh, Aryan Yazdan Parast, Mohammad Lotfollahi
arXiv Computer Vision
Oct 1

Looped Diffusion Transformer

arXiv:2609.40305v1 Announce Type: new Abstract: Improving text-to-image models has traditionally relied on increasing model size or the number of denoising steps. In this work, we explore an alternat...

By Yong Xien Chng, Tianyi Chen, Wenwen Tong, Haiwen Diao, Zhongang Cai, Lei Yang, Ziwei Liu, Lewei Lu, Dahua Lin, Gao Huang
arXiv AI
Sep 2

V-Co: A Closer Look at Visual Representation Alignment via Co-Denoising

V-Co investigates visual co-denoising for pixel-space diffusion models, using a unified JiT-based framework to isolate key design choices. The study identifies two essential components: a dual-stream architecture with flexible cross-stream interaction and a perceptual-drifting hybrid loss combined with RMS-based feature rescaling for stronger semantic supervision. Experiments on ImageNet-256 demonstrate that V-Co surpasses baseline pixel-space diffusion and strong prior pixel-diffusion methods at comparable model sizes while requiring fewer training epochs.

By Han Lin, Xichen Pan, Zun Wang, Yue Zhang, Chu Wang, Jaemin Cho, Mohit Bansal