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、[CL] Beneath the Surface of Consistency:Exploring Cross-lingual Knowledge Representation Sharing in LLMs 2、[LG] Learning Randomized Algorithms with Transformers 3、[CL] MagicDec:Breaking the Latency-Throughput Tradeoff for Long Context Generation with Speculative Decoding 4、[CL] Scaling Law with Learning Rate Annealing 5、[CL] To Code, or Not To Code? Exploring Impact of Code in Pre-training 摘要:探索LLM中的跨语言知识表示共享、用Transformer学习随机算法、通过推测式解码打破长上下文生成的延迟-吞吐量权衡、结合学习率退火的神经语言模型缩放率、探索代码数据在预训练中的正面影响 1、[CL] Beneath the Surface of Consistency: Exploring Cross-lingual Knowledge Representation Sharing in LLMs M Ifergan, L Choshen, R Aharoni, I Szpektor… [The Hebrew University of Jerusalem & Google Research] 一致性的
………………………………