爱可可 AI 前沿推介(3.5)

文章预览

LG - 机器学习 CV - 计算机视觉 CL - 计算与语言 1、[LG] Proteina：Scaling Flow-based Protein Structure Generative Models 2、[LG] Improve Representation for Imbalanced Regression through Geometric Constraints 3、[LG] How simple can you go? An off-the-shelf transformer approach to molecular dynamics 4、[LG] Steering Large Language Model Activations in Sparse Spaces 5、[LG] CoSMoEs：Compact Sparse Mixture of Experts 6、[LG] RSQ：Learning from Important Tokens Leads to Better Quantized LLMs 摘要：扩展基于流的蛋白质结构生成模型、通过几何约束改进不平衡回归的表征、分子动力学的“开箱即用“Transformer方法、在稀疏空间操控大语言模型激活、紧凑稀疏专家混合 1、[LG] Proteina: Scaling Flow-based Protein Structure Generative Models T Geffner, K Didi, Z Zhang, D Reidenbach... [NVIDIA] Proteina：扩展基于流的蛋白质结构生成模型 要点: Proteína：可扩展的基于 Flow 的蛋白质骨架生成 ………………………………