Learning to Deceive Knowledge Graph Augmented Models via Targeted Perturbation
arXiv:2010.12872
Abstract
Knowledge graphs (KGs) have helped neural models improve performance on various knowledge-intensive tasks, like question answering and item recommendation. By using attention over the KG, such KG-augmented models can also "explain" which KG information was most relevant for making a given prediction. In this paper, we question whether these models are really behaving as we expect. We show that, through a reinforcement learning policy (or even simple heuristics), one can produce deceptively perturbed KGs, which maintain the downstream performance of the original KG while significantly deviating from the original KG's semantics and structure. Our findings raise doubts about KG-augmented models' ability to reason about KG information and give sensible explanations.
arXiv comment13 pages, 11 figures
References in corpus (20)
- Semi-Supervised Classification with Graph Convolutional Networks
- Intensity and coherence of motifs in weighted complex networks
- Clustering in Complex Directed Networks
- Generalizations of the clustering coefficient to weighted complex networks
- Unifying Knowledge Graph Learning and Recommendation: Towards a Better Understanding of User Preferences
- Learning over Knowledge-Base Embeddings for Recommendation
- Learning Heterogeneous Knowledge Base Embeddings for Explainable Recommendation
- CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledge
- Neural Natural Language Inference Models Enhanced with External Knowledge
- Adversarial Attacks on Node Embeddings via Graph Poisoning
- Adversarial Attack on Graph Structured Data
- Fast Gradient Attack on Network Embedding
- Language Models as Knowledge Bases?
- Adversarial Examples on Graph Data: Deep Insights into Attack and Defense
- Attacking Graph Convolutional Networks via Rewiring
- A Survey of Adversarial Learning on Graphs
- Commonsense Knowledge Mining from Pretrained Models
- Scalable Multi-Hop Relational Reasoning for Knowledge-Aware Question Answering
- A Survey of Black-Box Adversarial Attacks on Computer Vision Models
- Exploiting Structured Knowledge in Text via Graph-Guided Representation Learning