Trick Me If You Can: Adversarial Writing of Trivia Challenge Questions

  title={Trick Me If You Can: Adversarial Writing of Trivia Challenge Questions},
  author={Eric Wallace and Pedro Rodriguez and Jordan L. Boyd-Graber},
Modern question answering systems have been touted as approaching human performance. However, existing question answering datasets are imperfect tests. Questions are written with humans in mind, not computers, and often do not properly expose model limitations. To address this, we develop an adversarial writing setting, where humans interact with trained models and try to break them. This annotation process yields a challenge set, which despite being easy for trivia players to answer… 

Figures and Tables from this paper

Quizbowl: The Case for Incremental Question Answering

This work makes two key contributions to machine learning research through Quizbowl: collecting and curating a large factoid QA dataset and an accompanying gameplay dataset, and developing a computational approach to playing Quiz Bowl that involves determining both what to answer and when to answer.

Adversarial NLI: A New Benchmark for Natural Language Understanding

This work introduces a new large-scale NLI benchmark dataset, collected via an iterative, adversarial human-and-model-in-the-loop procedure, and shows that non-expert annotators are successful at finding their weaknesses.

Towards Stronger Adversarial Baselines Through Human-AI Collaboration

This work presents a system that facilitates attack construction, combining human judgment with automated attacks to create better attacks more efficiently, and suggests that human-AI hybrid attacks are more effective than either human-only or AI-only attacks.

TruthfulQA: Measuring How Models Mimic Human Falsehoods

It is suggested that scaling up models alone is less promising for improving truthfulness than fine-tuning using training objectives other than imitation of text from the web.

What can AI do for me?: evaluating machine learning interpretations in cooperative play

This work designs a grounded, realistic human-computer cooperative setting using a question answering task, Quizbowl, and proposes an evaluation of interpretation on a real task with real human users, where the effectiveness of interpretation is measured by how much it improves human performance.

White-to-Black: Efficient Distillation of Black-Box Adversarial Attacks

This work trains a model to emulate the behavior of a white-box attack and shows that it generalizes well across examples, and training this model reduces adversarial example generation time by 19x-39x.

Generating Textual Adversarial Examples for Deep Learning Models: A Survey

This article reviews research works that address this difference and generate textual adversarial examples on DNNs and collects, select, summarize, discuss and analyze these works in a comprehensive way and cover all the related information to make the article self-contained.

Adversarial Attacks on Deep-learning Models in Natural Language Processing

A systematic survey on preliminary knowledge of NLP and related seminal works in computer vision is presented, which collects all related academic works since the first appearance in 2017 and analyzes 40 representative works in a comprehensive way.

QA Dataset Explosion: A Taxonomy of NLP Resources for Question Answering and Reading Comprehension

This study is the largest survey of the deep learning models in NLP to date, providing an overview of the various formats and domains of the current resources, and highlighting the current lacunae for future work.

Learning What Makes a Difference from Counterfactual Examples and Gradient Supervision

This work proposes an auxiliary training objective that improves the generalization capabilities of neural networks by leveraging an overlooked supervisory signal found in existing datasets, a.k.a counterfactual examples, which provide a signal indicative of the underlying causal structure of the task.



Adversarial Examples for Evaluating Reading Comprehension Systems

This work proposes an adversarial evaluation scheme for the Stanford Question Answering Dataset that tests whether systems can answer questions about paragraphs that contain adversarially inserted sentences without changing the correct answer or misleading humans.

SQuAD: 100,000+ Questions for Machine Comprehension of Text

A strong logistic regression model is built, which achieves an F1 score of 51.0%, a significant improvement over a simple baseline (20%).

Semantically Equivalent Adversarial Rules for Debugging NLP models

This work presents semantically equivalent adversaries (SEAs) – semantic-preserving perturbations that induce changes in the model’s predictions that induce adversaries on many instances that are extremely similar semantically.

Studio Ousia's Quiz Bowl Question Answering System

In this chapter, we describe our question answering system, which was the winning system at the Human–Computer Question Answering (HCQA) Competition at the Thirty-first Annual Conference on Neural

Reading Wikipedia to Answer Open-Domain Questions

This approach combines a search component based on bigram hashing and TF-IDF matching with a multi-layer recurrent neural network model trained to detect answers in Wikipedia paragraphs, indicating that both modules are highly competitive with respect to existing counterparts.

Annotation Artifacts in Natural Language Inference Data

It is shown that a simple text categorization model can correctly classify the hypothesis alone in about 67% of SNLI and 53% of MultiNLI, and that specific linguistic phenomena such as negation and vagueness are highly correlated with certain inference classes.

Question Answering with Subgraph Embeddings

A system which learns to answer questions on a broad range of topics from a knowledge base using few hand-crafted features, using low-dimensional embeddings of words and knowledge base constituents to score natural language questions against candidate answers.

Human-Computer Question Answering: The Case for Quizbowl

The setting: the game of quiz bowl is described, it is argued why it makes a suitable game for human-computer competition, and the logistics and preparation for the competition are described.

A Thorough Examination of the CNN/Daily Mail Reading Comprehension Task

A thorough examination of this new reading comprehension task by creating over a million training examples by pairing CNN and Daily Mail news articles with their summarized bullet points, and showing that a neural network can be trained to give good performance on this task.

Stress Test Evaluation for Natural Language Inference

This work proposes an evaluation methodology consisting of automatically constructed “stress tests” that allow us to examine whether systems have the ability to make real inferential decisions, and reveals strengths and weaknesses of these models with respect to challenging linguistic phenomena.