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 1、[LG] Towards an Improved Understanding and Utilization of Maximum Manifold Capacity Representations 2、[AS] Joint Audio and Symbolic Conditioning for Temporally Controlled Text-to-Music Generation 3、[LG] Refusal in Language Models Is Mediated by a Single Direction 4、[CV] VoCo-LLaMA：Towards Vision Compression with Large Language Models 5、[CL] A Simple and Effective L2 Norm-Based Strategy for KV Cache Compression 摘要：更好地理解和利用MMCR方法、联合音频和符号调节实现时序控制的文本到音乐的生成、语言模型中的拒绝行为可用单一方向进行调节、利用大型语言模型实现视觉压缩、简单有效的基于L2范数的KV缓存压缩策略 1、[LG] Towards an Improved Understanding and Utilization of Maximum Manifold Capacity Representations R Schaeffer, V Lecomte, D B Pai, A Carranza… [Stanford CS] 更好地理解和利用MMCR方法 要点:
………………………………
