爱可可 AI 前沿推介(5.28)

文章预览

LG - 机器学习 CV - 计算机视觉 CL - 计算与语言 1、[LG] Denoising LM：Pushing the Limits of Error Correction Models for Speech Recognition 2、[LG] The Impact of Geometric Complexity on Neural Collapse in Transfer Learning 3、[LG] The Road Less Scheduled 4、[LG] Automatic Data Curation for Self-Supervised Learning： A Clustering-Based Approach 5、[LG] Grokked Transformers are Implicit Reasoners：A Mechanistic Journey to the Edge of Generalization 摘要：用去噪语言模型探索语音识别错误校正模型的极限、迁移学习中几何复杂度对神经坍缩的影响、"无时间表"学习方法、面向自监督学习的基于聚类的自动数据整理、通过“Grokking”训练的Transformer是隐式推理器 1、[LG] Denoising LM: Pushing the Limits of Error Correction Models for Speech Recognition Z Gu, T Likhomanenko, H Bai, E McDermott… [Apple] 用去噪语言模型探索语音识别错误校正模型的极限 要点: 去噪语言模型(DLM)是 ………………………………