老牌Transformer杀手在ICLR悄然更新：Mamba-3三大改进趋近设计完全体
"长序列处理" 相关文章
LouisKV: Efficient KV Cache Retrieval for Long Input-Output Sequences
Native Hybrid Attention for Efficient Sequence Modeling
Recurrence-Complete Frame-based Action Models
Beyond Next-Token Prediction: A Performance Characterization of Diffusion versus Autoregressive Language Models
Replacing Softmax Similarity with a Sharpened Angular Similarity: Theory and Practice of Scaling To Billion-Context Attention
InfLLM-V2: Dense-Sparse Switchable Attention for Seamless Short-to-Long Adaptation
SCOUT: Toward Sub-Quadratic Attention via Segment Compression for Optimized Utility in Transformers
Transformer的并行计算与长序列处理瓶颈
ROVER: Recursive Reasoning Over Videos with Vision-Language Models for Embodied Tasks
首次将单目3D重建推向公里级极限！南开、南大提出VGGT-Long：分块、循环、对齐，开源
Scaling Attention to Very Long Sequences in Linear Time with Wavelet-Enhanced Random Spectral Attention (WERSA)
算力终结者来了，华人天团「降维打击」注意力瓶颈，AI狂飙进对数时代
Deepseek NSA可能是 Transfermer 的新解法
