文章预览
LG - 机器学习 CV - 计算机视觉 CL - 计算与语言 AS - 音频与语音 RO - 机器人 1、[LG] Geometry of Lightning Self-Attention：Identifiability and Dimension 2、[LG] Beyond Preferences in AI Alignment 3、[RO] Bidirectional Decoding：Improving Action Chunking via Closed-Loop Resampling 4、[LG] Training Ultra Long Context Language Model with Fully Pipelined Distributed Transformer 5、[CL] CLOCR-C：Context Leveraging OCR Correction with Pre-trained Language Models 摘要：闪电自注意力的几何学、超越偏好的人工智能对齐、通过闭环重采样改进动作分块、用全流水线分布式Transformer训练超长上下文语言模型、利用预训练语言模型进行上下文关联OCR校正 1、[LG] Geometry of Lightning Self-Attention: Identifiability and Dimension N W. Henry, G L Marchetti, K Kohn [University of Toronto & Royal Institute of Technology (KTH)] 闪电自注意力的几何学：可识别性和维度 要点： 本文利用代数几
………………………………
