# Trick Me If You Can: Human-in-the-loop Generation of Adversarial Examples for Question Answering

arXiv: https://arxiv.org/abs/1809.02701 · HTML: https://mlsanity.com/paper/1809.02701/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 7 Sep 2018.

## Built on by (most cited first, 6 of 6)

- [TruthfulQA: Measuring How Models Mimic Human Falsehoods](https://mlsanity.com/paper/2109.07958/) (2021)
- [QA Dataset Explosion: A Taxonomy of NLP Resources for Question Answering and Reading Comprehension](https://mlsanity.com/paper/2107.12708/) (2021)
- [Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning Models](https://mlsanity.com/paper/2503.01781/) (2025)
- [SUCEA: Reasoning-Intensive Retrieval for Adversarial Fact-checking through Claim Decomposition and Editing](https://mlsanity.com/paper/2506.04583/) (2025)
- [From Seed to Harvest: Augmenting Human Creativity with AI for Red-teaming Text-to-Image Models](https://mlsanity.com/paper/2507.17922/) (2025)
- [Do LLMs and Humans Find the Same Questions Difficult? A Case Study on Japanese Quiz Answering](https://mlsanity.com/paper/2511.12300/) (2025)

## Abstract (arXiv)

Adversarial evaluation stress tests a model's understanding of natural language. While past approaches expose superficial patterns, the resulting adversarial examples are limited in complexity and diversity. We propose human-in-the-loop adversarial generation, where human authors are guided to break models. We aid the authors with interpretations of model predictions through an interactive user interface. We apply this generation framework to a question answering task called Quizbowl, where trivia enthusiasts craft adversarial questions. The resulting questions are validated via live human--computer matches: although the questions appear ordinary to humans, they systematically stump neural and information retrieval models. The adversarial questions cover diverse phenomena from multi-hop reasoning to entity type distractors, exposing open challenges in robust question answering.