作者位次 #6/6ICLR 2026Poster· Computer Vision / Vision Models & Multimodal
Reversible Primitive–Composition Alignment for Continual Vision–Language Learning
Canran Xiao · Tianxiang Xu · SiYuan Ma · Yiyang Jiang · Haoyu Gao · Yuhan Wu
首次获取该候选人的完整档案及 StartLux 评估共 20 积分；每位用户单独解锁，之后查看免费。主动刷新 20 积分。
实时查找电子邮箱，费用不超过20积分；未找到邮箱不扣费。
Canran Xiao · Tianxiang Xu · SiYuan Ma · Yiyang Jiang · Haoyu Gao · Yuhan Wu
Yuhua Zhou · Shichao Weng · Changhai Zhou · Yuhan Wu · Qian Qiao · Jun Gao · Fei Yang · Aimin Pan
While the massive scale of modern LLMs enables remarkable performance, their static, input-agnostic computational graph incurs substantial resource wastage and high latency during inference. Existing dynamic schemes, such as early-exit and layer-drop reduce FLOPs but break batch processing or int...