arXiv cs.RO· Qize Yu, Lianrui Fan, Bowen Ping, Xini Ding, Zetian Song, Junbo Niu, Kaixuan Wang, Tianxing Chen, Yue Chen, Minghua He, Yuran Wang, Jie Huang, Haojun Zhang, Min Chen, Hao Li, Wenxuan Song, Ruihai Wu, Xianming Liu, Shilong Liu, Shuchang Zhou, Ping Luo, Shiyu Huang·· 4 小时前AI 评分57
GroundAnything：用块状去噪兼顾并行解码与精确视觉定位
GroundAnything: Reconciling Parallel Decoding with Precise Visual Grounding at Flash Speed
AI 导读
GroundAnything 是一个 40 亿参数的视觉定位基础模型，通过块状去噪把并行解码与精确目标定位结合，训练融合公开数据集预训练、自回归到扩散的直接转换、监督微调和基于 GRPO 的强化后训练。
来源：arXiv cs.RO · arxiv.org