Publication Type

Journal Article

Version

acceptedVersion

Publication Date

4-2021

Abstract

The ability to read, reason, and infer lies at the heart of neural reasoning architectures. After all, the ability to perform logical reasoning over language remains a coveted goal of Artificial Intelligence. To this end, models such as the Turing-complete differentiable neural computer (DNC) boast of real logical reasoning capabilities, along with the ability to reason beyond simple surface-level matching. In this brief, we propose the first probe into DNC's logical reasoning capabilities with a focus on text-based question answering (QA). More concretely, we propose a conceptually simple but effective adversarial attack based on metamorphic relations. Our proposed adversarial attack reduces DNCs' state-of-the-art accuracy from 100% to 1.5% in the worst case, exposing weaknesses and susceptibilities in modern neural reasoning architectures. We further empirically explore possibilities to defend against such attacks and demonstrate the utility of our adversarial framework as a simple scalable method to improve model adversarial robustness.

Keywords

Task analysism, Cognition, Plugs;Perturbation methods, Memory modules, Computer architecture, Computational modeling, Adversarial examples, deep learning, differentiable neural computer (DNC), supervised learning

Discipline

OS and Networks | Software Engineering

Research Areas

Software and Cyber-Physical Systems

Publication

IEEE Transactions on Neural Networks and Learning Systems

First Page

1

Last Page

7

ISSN

2162-2388

Identifier

10.1109/TNNLS.2021.3072166

Publisher

Institute of Electrical and Electronics Engineers

Download

Share

COinS