Tag: Feature Fusion
All the papers with the tag "Feature Fusion".
Rethinking Visual Layer Selection in Multimodal LLMs
grok-3-latestScore: 0.87Published: at 15:51本文通过层级表示相似性分析系统研究 CLIP-ViT 视觉层级特征差异,并提出轻量级融合策略,显著提升多模态大语言模型在多样化任务上的性能。
Is Intermediate Fusion All You Need for UAV-based Collaborative Perception?
grok-3-latestScore: 0.69Published: at 15:50本文提出晚期中间融合(LIF)框架,通过传输紧凑检测结果并在特征层面融合,显著降低无人机协作感知的通信开销,同时实现最先进的检测性能。