Tencent（腾讯） - arXivDaily 大厂专区

2607.08497 2026-07-10 cs.CV cs.AI cs.CL cs.LG 新提交

Cognitive-structured Multimodal Agent for Multimodal Understanding, Generation, and Editing

用于多模态理解、生成和编辑的认知结构多模态智能体

Feng Wang, Canmiao Fu, Zhipeng Huang, Chen Li, Jing Lyu, Ge Li

机构 * Peking University（北京大学）； Tencent Inc（腾讯公司）

AI总结研究针对统一多模态模型在长时多模态对话中的局限，提出认知结构多模态智能体，通过外化视觉信息等方式改进。开发统一场景引擎，构建评估基准。实验显示该智能体检索准确率高、推理时间减半，还介绍了工具包，为长时多模态智能体提供新范式。

Comments 16 pages, 7 figures, 8 tables. Project page: https://caseclose.github.io/cma-harness/ Code: https://github.com/caseclose/cma-harness

2607.08186 2026-07-10 cs.CL 新提交

Hidden Decoding at Scale: Latent Computation Scaling for Large Language Models

大规模隐藏解码：大语言模型的潜在计算扩展

Aiwei Liu, Cheng Shi, Chuhan Wu, Ci Lei, Di Lu, Donald He, Fan Zhang, Fanhao Kong, Feifei Zhang, Guan Wang, Haicheng Wang, Haoyu Liu, Houjin Yu, Jiachen Ding, Jiayi Feng, Jie Zhou, Jijun Chi, Jindi Shi, Jing Lei, Junjie Zhang, Laiyi Li, Le Tian, Linhao Zhang, Miao Fan, Sijun Zhang, Wei Jia, Weiwei Shi, Wenhan Li, Wentao Zhao, Wenteng Liang, Xiao Zhou, Xiaojin Zhou, Xihuai Wang, Xinyu Gao, Xuanliang Wang, Xuyang Ao, Yang Yu, Yangxiu You, Yinuo Zhao, Yufei Kuang, Yufei Wang, Yuan Liu, Yuan Liu, Yuwen Chen, Zhencong Tian, Zhongyin Zhao, Zilin Yu, Zitao Wang

机构 * WeChat AI Team（微信人工智能团队）

AI总结研究在Transformer主干固定时，通过为token分配更多计算改进大语言模型，提出隐藏解码方法，将token扩展为n个流并结合流分解注意力，实验验证该方法在前沿规模的有效性及跨扩展因子时收益随n增加。

Comments 30 pages, 9 figures