NeoMME is a family of 260M and 800M‑parameter multimodal‑native multilingual encoders that process text and raw image patches in a single bidirectional Transformer. Trained from scratch with a masked discrete‑diffusion objective conditioned on visible image patches, NeoMME supports a 16,384‑token context, enabling encoding of up to two 4K UHD images. In downstream tests, NeoMME‑Retriever models outperform all sub‑800M‑parameter baselines on the ViDoRe v3 benchmark and achieve twice the throughput of ColModernVBERT on an NVIDIA L40S, while hierarchical token pooling and asymmetric quantization compress embeddings 255× with minimal loss in retrieval performance.
By Aur\'elien Lac, Tony Wu
JUICE (Joint Unified multimodal modeling In Continuous Encoder spaces) is a single flow‑matching Transformer that jointly models images and text in pretrained continuous encoder spaces, allowing each modality to be generated by the same process with its own noise level. The model can perform tasks such as image generation, captioning, visual question answering, and joint generation by deciding which modalities are clean and which are generated. Experiments show that joint and conditional denoising together outperform either alone, and that JUICE achieves strong performance on GenEval, VQAv2, and visual spatial planning tasks, even enabling coupled sampling where both modalities are generated simultaneously.
By Jiahan Li, Huangyuan Su, Shengbang Tong, Shusheng Yang, Ruihan Guo, Suozhi Huang, Hexu Zhao, Pratyusha Sharma, Sherry Yang, Jiatao Gu, Rob Fergus, Kyunghyun Cho
arXiv:2609.40362v1 Announce Type: new
Abstract: We present Multimodal Flow, a fully continuous generative model of language and vision. Most unified multimodal models either model both language and q...
By Hongyuan Tao, Xinggang Wang, Lianghui Zhu, Yongkang Li, Yunchao Wei, Bin Feng, Shaoyu Chen, Qian Zhang, Chang Huang, Kai Yu
Intern Lumina U2 is a diffusion large language model that uses multi-codebook visual tokens in a fully discrete space, allowing each visual position to be represented by a group of complementary codes. The model concatenates embeddings at each position and employs a lightweight autoregressive head to predict codes sequentially while the diffusion backbone processes all positions in parallel. Built on the unified 4D representation of AToken, it supports image, video, and 3D inputs, achieves leading performance in visual understanding, image generation, and editing, and includes a cross‑platform training, inference, and evaluation system with a stability‑aware sampling strategy that speeds up image generation by 4.45×.
By Qian Yu, Zongyang Qiu, Juncheng Yan, Hongliang Li, Qi Qin, Keqi Wang, Siqi Luo, Junkui Ju, Yapeng Mi, Tiantian Lu, Tian Liu, Pengyu Xie, Chenhao Ji, Minghui Zhang, Yiang Zhao, Siyu Yan, Haiying Xu, Yanbo Wang, Jianghan Shen, Yuandong Pu, Jiayi Lei, Borui Zhang, Jinxin Shi, Bo Zhang, Yuhang Zang, Shenglong Ye, Lixin Gu, Hongjie Zhang, Linghan Liu, Xinyuan Chen, Yuewen Cao, Lihe Zhang, Xiaohong Liu, Junjun He, Qiangwei Liu, Qi Zhang, Yu Cheng, Tianfan Xue, Hui Xiong, Chang Xu, Yu Qiao, Yihao Liu, Bin Fu
arXiv:2607. 25527v1 Announce Type: cross Abstract: Unifying visual understanding and generation in one model holds immense promise, but remains challenging and expensive due to heavy compute and data demands and conflicts between the visual features needed for these two capabilities.
By Weiming Zhuang, Jiabo Huang, Jingtao Li, Zhizhong Li, Chen Chen, Sina Sajadmanesh, Lingjuan Lyu
VisionWeave introduces elastic visual representation weaving, a native capability for multimodal large language models that learns where and at what granularity to encode visual information. The method combines a gated spatial pooler for coarse representations with a granularity router that allocates content‑adaptive token usage, trained end‑to‑end on large‑scale data. Experiments on Qwen3.5‑4B and Qwen3.8‑27B show that VisionWeave can save 43.0% of tokens while preserving 98.9% of performance across eight benchmarks, and delivers significant throughput gains and latency reductions when deployed on the SGLang serving engine.
By Yuan Feng, Qize Yang, Ruizhe Chen, Sibo Song, Haolin He, Muzhi Zhu, Zihan Liu, Yunfei Chu, Xize Cheng, Yuxuan Wang, Jin Xu, Xike Xie
Unified multimodal generation covers discrete, continuous, and hybrid paradigms. Discrete approaches with single-codebook visual tokenizers align naturally with text modeling but have limited represen...
arXiv:2609.37225v1 Announce Type: cross
Abstract: Multimodal large language models (MLLMs) have shown strong potential for universal multimodal representation learning. However, existing methods eith...
By Zijing Cai, Yuzhe Wang, Jingxian Zhu, Fengbin Zhu, Richang Hong
TokenSwap introduces a method to benchmark and reduce the modality gap in multimodal large language models (MLLMs) by interleaving semantically aligned images with text tokens. By transforming text-only benchmarks like MMLU into image‑interleaved versions (TokenSwap‑Bench), the authors reveal a pervasive performance drop—averaging 19.6%—across 42 MLLMs, with reasoning models showing smaller gaps. Training with TokenSwap mitigates this gap while maintaining strong text‑only and vision‑language performance.
By Andong Hua, Colton Bishop, Igor Mordatch, Arian Hosseini, Jindong Gu, Aleksandra Faust, Rebecca Roelofs, Yao Qin
arXiv:2606. 09131v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) commonly inherit the deep, symmetric Transformer backbone designed for unimodal text modeling, and apply the same computation uniformly to image and language tokens.
By Siyuan Liu, Jinyang Wu
arXiv:2606. 13289v1 Announce Type: cross Abstract: Holistic visual tokenizers are fundamental to unified multimodal models (UMMs) as they map diverse visual inputs into a unified representation space.
By Guozhen Zhang, Xuerui Qiu, Yutao Cui, Tianhui Song, Changlin Li, Junzhe Li, Tao Huang, Xiao Zhang, Yang Li, Jianbing Wu, Miles Yang, Zhao Zhong, Liefeng Bo, Limin Wang
Vision-language models commonly project all tokens produced by a pretrained vision encoder into a large language model. However, final-layer features can discard text, local attributes, and spatial relationships, while high-resolution inputs substantially increase context length and inference latency.