LARC (Language-annotated Abstraction and Reasoning)

LARC是从ARC (抽象和推理语料库) 构建的数据集。ARC是一组任务，用于测试代理灵活解决新问题的能力。虽然大多数ARC任务对人类来说都很容易，但对最先进的人工智能来说却是一个挑战。
LARC或带有语言注释的ARC是一组人类参与者的自然语言描述的集合，这些参与者既不熟悉ARC又彼此不熟悉，他们互相指导如何解决ARC任务。LARC包含88% ARC任务的成功说明。

LARC is a dataset constructed from ARC (Abstract and Reasoning Corpus). ARC is a collection of tasks designed to test agents' ability to flexibly solve novel problems. While most ARC tasks are straightforward for humans, they pose a significant challenge to state-of-the-art artificial intelligence systems. LARC, also referred to as ARC with Linguistic Annotations, is a corpus of natural language descriptions produced by human participants who were both unfamiliar with ARC and mutually unacquainted, as they instructed one another on how to solve ARC tasks. LARC contains successful explanatory walkthroughs for 88% of ARC tasks.

搜集汇总

背景与挑战

背景概述

LARC是一个基于ARC构建的数据集，通过收集人类参与者的自然语言描述来指导解决ARC任务，覆盖了88%的ARC任务成功说明。该数据集由麻省理工学院和Autodesk Research于2021年发布，旨在测试人工智能解决新问题的能力。