bAbI

bAbI 数据集是由 20 个不同任务组成的文本 QA 基准。每项任务都旨在测试不同的推理技能，例如演绎、归纳和共指解析。一些任务需要关系推理，例如，比较不同实体的大小。每个样本由一个问题、一个答案和一组事实组成。该数据集有两个版本，指的是不同的数据集大小：bAbI-1k 和 bAbI-10k。 bAbI-10k 版本的数据集由每个任务 10,000 个训练样本组成。

资料来源：工作记忆网络：使用关系推理模块增强记忆网络

The bAbI dataset is a textual question answering (QA) benchmark composed of 20 distinct tasks. Each task is designed to test diverse reasoning skills, including deductive reasoning, inductive reasoning, and coreference resolution. Certain tasks require relational reasoning, such as comparing the sizes of different entities. Each sample consists of a question, an answer, and a set of supporting facts. The dataset offers two variants differentiated by dataset scale: bAbI-1k and bAbI-10k. The bAbI-10k variant includes 10,000 training samples for each task. Source: Working Memory Networks: Augmenting Memory Networks with a Relational Reasoning Module

搜集汇总

背景与挑战

背景概述

bAbI是一个包含20个不同任务的文本问答基准数据集，旨在测试多种推理技能，如演绎、归纳和共指解析。该数据集有两个版本(bAbI-1k和bAbI-10k)，其中bAbI-10k版本每个任务包含10,000个训练样本。