Unified Multimodal Autoregressive Modeling with Shared Context-Visual Tokenizer is Key to Unification
Summary
UniAR presents a unified autoregressive framework that uses a single discrete visual tokenizer to bridge visual understanding and generation, achieving state-of-the-art results in image generation and editing.
View Cached Full Text
Cached at: 06/17/26, 03:36 AM
Paper page - Unified Multimodal Autoregressive Modeling with Shared Context-Visual Tokenizer is Key to Unification
Source: https://huggingface.co/papers/2606.18249
Abstract
UniAR presents a unified autoregressive framework that uses a single discrete visual tokenizer to bridge visual understanding and generation, achieving state-of-the-art results in image generation and editing through multi-level feature fusion, bitwise quantization, and parallel prediction.
Unified Multimodal Modeling aims to integrate visual understanding and generation within a single system. However, existing approaches typically rely on two disparate visual tokenizers, which splits the representation space and hinders truly unified modeling. We propose UniAR, aunified autoregressive frameworkwhere a singlediscrete visual tokenizerserves as the key bridge between understanding and generation, enabling a shared context in which the model can directly interpret its own generated visual tokens without additional re-encoding. UniAR adapts a pretrained vision encoder withmulti-level feature fusionand alookup-free bitwise quantizationscheme, preserving both high-level semantics and low-level details while scaling the effective visual vocabulary at minimal cost. Building on this, the unified autoregressive model adoptsparallel-bitwise-predictionto jointly predict spatially grouped, multi-level visual codes, substantially reducingvisual sequence lengthand accelerating generation. Finally, adiffusion-based visual decoderoperates on discrete visual tokens to decode high-fidelity images. Through large-scale pre-training, followed bysupervised fine-tuningandreinforcement learning, UniAR achieves state-of-the-art performance on image generation and image editing while remaining competitive on multimodal understanding benchmarks. The project page is available at https://sharelab-sii.github.io/uniar-web.
View arXiv pageView PDFProject pageGitHub2Add to collection
Models citing this paper2
#### ShareLab-SII/UniAR-RL Image-to-Text• 10B• Updatedabout 2 hours ago
#### ShareLab-SII/UniAR-SFT Image-to-Text• 10B• Updatedabout 2 hours ago
Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2606.18249 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2606.18249 in a Space README.md to link it from this page.
Collections including this paper0
No Collection including this paper
Add this paper to acollectionto link it from this page.
Similar Articles
Studying Image Tokenizers as Visual Languages in Unified Multimodal Models
This paper analyzes how image tokenizer design affects joint text-image modeling in multimodal models using a controlled autoregressive testbed, showing distinct scaling behaviors and correlations with downstream performance.
ARM: An AutoRegressive Large Multimodal Model with Unified Discrete Representations
ARM presents a unified autoregressive framework for image understanding, generation, and editing using discrete semantic tokenization and reinforcement learning optimization, showing cross-task synergy.
HiUni: Harmonizing Understanding and Generation in Shared Visual Space
HiUni explores extending a pretrained vision-language model into a unified multimodal model via extensive parameter sharing, reusing a single dense backbone for both understanding and generation with a PixelDiffusionHead and HiGate pathway to address optimization and spatial-detail challenges. Experiments show it matches the pretrained VLM's understanding score while achieving competitive image generation and editing performance.
Steering Visual Generation in Unified Multimodal Models with Understanding Supervision
This paper introduces UNO, an Understanding-Oriented Post-Training framework that uses comprehension tasks as supervisory signals to enhance image generation and editing in unified multimodal models.
UniCorn: Towards Self-Improving Unified Multimodal Models through Self-Generated Supervision
UniCorn is a framework that enables unified multimodal models to self-improve by using a multi-agent system for prompt generation, image creation, and quality evaluation, achieving state-of-the-art results on text-to-image benchmarks like TIIF, WISE, and OneIG-EN.