Language instruction plays an essential role in the natural language groundednavigation tasks . We propose a DynamicReinforced Instruction Attacker (DR-Attacker) which learns to mislead thenavigator to move to the wrong target by destroying the most instructive information in instructions at different timesteps . DR-Attacker is optimized by the reinforcement learning algorithm to generate perturbed instructions sequentially during the navigation, according to a learnable attack score . The proposed method is overstate-of-the-art and shows the superiority of our proposed method over state-of the-art methods . Code is available at https://://://github.com/expectorlin/DR-attacker.code is available in the source of the proposed DR-ATTACKER and the code is available to download at http://http://www.gene-report-to-da-da.org/dna-da/dana-dna.

Author(s) : Bingqian Lin, Yi Zhu, Yanxin Long, Xiaodan Liang, Qixiang Ye, Liang Lin

Links : PDF - Abstract

Code :
Coursera

Keywords : attacker - dr - proposed - da - instruction -

Leave a Reply

Your email address will not be published. Required fields are marked *