Continual-ARC: Measuring Skill Acquisition over a Working Lifetime

媒体报道 Hugging Face Papers 原文语言:中文

Continual-ARC是一个由ARC-AGI任务组成的终身流,用于在冻结权重的智能体框架下比较基于记忆和基于权重的持续学习。研究表明,框架的持久性很少能改善记忆,大多数框架方法的表现不及简单的缓存,而基于权重的学习则表现出更好的泛化能力。

全文

Continual-ARC, a lifelong stream of ARC-AGI tasks, compares memory- and weight-based continual learning under frozen-weight agent harnesses, revealing that harness persistence rarely improves memorization and most harness methods underperform simple caching, while weight-based learning shows better generalization.

正文以来源语言显示。

参考来源

← 返回每日AI报告