BRAINTEASER: Lateral Thinking Puzzles for Large Language Models
Preprint
Abstract
The success of language models has inspired the NLP community to attend to tasks that require implicit and complex reasoning, relying on human-like commonsense mechanisms. While such vertical thinking tasks have been relatively popular, lateral thinking puzzles have received little attention. To bridge this gap, we devise BRAINTEASER: a multiple-choice Question Answering task designed to test the model's ability to exhibit lateral thinking and defy default commonsense associations. We design a three-step procedure for creating the first lateral thinking benchmark, consisting of data collection, distractor generation, and generation of adversarial examples, leading to 1,100 puzzles with high-quality annotations. To assess the consistency of lateral reasoning by models, we enrich BRAINTEASER based on a semantic and contextual reconstruction of its questions. Our experiments with state-of-the-art instruction- and commonsense language models reveal a significant gap between human and model performance, which is further widened when consistency across adversarial formats is considered. We make all of our code and data available to stimulate work on developing and evaluating lateral thinking models.Keywords
All Related Versions
This publication has 21 references indexed in Scilit:
- Back to Square One: Artifact Detection, Training and Commonsense Disentanglement in the Winograd Schema2021
- Shortcutted Commonsense: Data Spuriousness in Deep Learning of Commonsense Reasoning2021
- CommonGen: A Constrained Text Generation Challenge for Generative Commonsense Reasoning2020
- ProtoQA: A Question Answering Dataset for Prototypical Common-Sense Reasoning2020
- Adversarial NLI: A New Benchmark for Natural Language Understanding2020
- The Sensitivity of Language Models and Humans to Winograd Schema Perturbations2020
- Dark Motives and Elective Use of Brainteaser Interview QuestionsApplied Psychology, 2018
- Learning psychology from riddles: The case of stumpersJudgment and Decision Making, 2018
- Catalytic assessment: understanding how MCQs and EVS can foster deep learningBritish Journal of Educational Technology, 2009
- Lateral Thinking and Technology EducationJournal of Science Education and Technology, 1997