TY - GEN
T1 - CheersA large glass mug filled with dark beer topped with a thick, foamy head. The mug has a handle and condensation on the outside, suggesting the beer is cold. The background is plain white, emphasizing the rich color and texture of the beer and foam. The image highlights the refreshing and appealing qualities of a freshly poured dark beer.
T2 - 19th European Conference on Computer Vision, ECCV 2026
AU - Zhang, Yichen
AU - Peng, Da
AU - Guo, Zonghao
AU - Zhang, Zijian
AU - Yang, Xuesong
AU - Sun, Tong
AU - Sun, Shichu
AU - Zhang, Yidan
AU - Li, Yanghao
AU - Zhao, Haiyan
AU - Xu, Wang
AU - Shi, Qi
AU - Sun, Yangang
AU - Chen, Chi
AU - Wang, Shuo
AU - Yan, Yukun
AU - Han, Xu
AU - Ma, Qiang
AU - Ke, Wei
AU - Wang, Liang
AU - Liu, Zhiyuan
AU - Sun, Maosong
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.
PY - 2026
Y1 - 2026
N2 - A recent cutting-edge topic in multimodal modeling is to unify visual comprehension and generation within a single model. However, the two tasks demand mismatched decoding regimes and visual representations, making it non-trivial to jointly optimize within a shared feature space. In this work, we present Cheers, a unified multimodal model that decouples patch-level details from semantic representations, thereby stabilizing semantics for multimodal understanding and improving fidelity for image generation via gated detail residuals. Cheers includes three key components: (i) a unified vision tokenizer that encodes and compresses image latent states into semantic tokens for efficient LLM conditioning, (ii) an LLM-based Transformer that unifies autoregressive decoding for text generation and diffusion decoding for image generation, and (iii) a cascaded flow matching head that decodes visual semantics first and then injects semantically gated detail residuals from the vision tokenizer to refine high-frequency content. Experiments on popular benchmarks demonstrate that Cheers matches or surpasses advanced UMMs in both visual understanding and generation. Notably, Cheers outperforms the Tar-1.5B on the popular benchmarks GenEval and MMBench, while requiring only 20% of the training cost, indicating effective and efficient (i.e., 4× token compression) unified multimodal modeling. Code and checkpoints are available at github.com/AI9Stars/CHEERS.
AB - A recent cutting-edge topic in multimodal modeling is to unify visual comprehension and generation within a single model. However, the two tasks demand mismatched decoding regimes and visual representations, making it non-trivial to jointly optimize within a shared feature space. In this work, we present Cheers, a unified multimodal model that decouples patch-level details from semantic representations, thereby stabilizing semantics for multimodal understanding and improving fidelity for image generation via gated detail residuals. Cheers includes three key components: (i) a unified vision tokenizer that encodes and compresses image latent states into semantic tokens for efficient LLM conditioning, (ii) an LLM-based Transformer that unifies autoregressive decoding for text generation and diffusion decoding for image generation, and (iii) a cascaded flow matching head that decodes visual semantics first and then injects semantically gated detail residuals from the vision tokenizer to refine high-frequency content. Experiments on popular benchmarks demonstrate that Cheers matches or surpasses advanced UMMs in both visual understanding and generation. Notably, Cheers outperforms the Tar-1.5B on the popular benchmarks GenEval and MMBench, while requiring only 20% of the training cost, indicating effective and efficient (i.e., 4× token compression) unified multimodal modeling. Code and checkpoints are available at github.com/AI9Stars/CHEERS.
KW - Unified multimodal model
KW - Unified vision encoder
KW - Visual generation and comprehension
UR - https://www.scopus.com/pages/publications/105052175420
U2 - 10.1007/978-3-032-37098-3_22
DO - 10.1007/978-3-032-37098-3_22
M3 - 会议稿件
AN - SCOPUS:105052175420
SN - 9783032370976
T3 - Lecture Notes in Computer Science
SP - 398
EP - 417
BT - Computer Vision – ECCV 2026 - 19th European Conference, Malmö, Sweden, September 8–12, 2026, Proceedings
A2 - Favaro, Paolo
A2 - Kukelova, Zuzana
A2 - Maki, Atsuto
A2 - Rohrbach, Anna
A2 - Schindler, Konrad
A2 - Tombari, Federico
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 8 September 2026 through 12 September 2026
ER -