BriefGPT.xyz
Nov, 2017
神经过程网络模拟行为动态
Simulating Action Dynamics with Neural Process Networks
HTML
PDF
Antoine Bosselut, Omer Levy, Ari Holtzman, Corin Ennis, Dieter Fox...
TL;DR
本文提出了神经过程网络作为理解过程文本的一种新方法，通过动态实体跟踪和显式地建模动作作为状态转换器，实现对于行为的动态建模来理解和生成程序文本，并且从中提供更准确的背景信息和可解释的内部表示。
Abstract
Understanding
procedural language
requires anticipating the
causal effects
of actions, even when they are not explicitly stated. In this work, we introduce
→