REBUS: A Robust Evaluation Benchmark of Understanding Symbols

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Gritsevskiy, Andrew, Panickssery, Arjun, Kirtland, Aaron, Kauffman, Derik, Gundlach, Hans, Gritsevskaya, Irina, Cavanagh, Joe, Chiang, Jonathan, La Roux, Lydia, Hung, Michelle
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916272123412480
author Gritsevskiy, Andrew
Panickssery, Arjun
Kirtland, Aaron
Kauffman, Derik
Gundlach, Hans
Gritsevskaya, Irina
Cavanagh, Joe
Chiang, Jonathan
La Roux, Lydia
Hung, Michelle
author_facet Gritsevskiy, Andrew
Panickssery, Arjun
Kirtland, Aaron
Kauffman, Derik
Gundlach, Hans
Gritsevskaya, Irina
Cavanagh, Joe
Chiang, Jonathan
La Roux, Lydia
Hung, Michelle
contents We propose a new benchmark evaluating the performance of multimodal large language models on rebus puzzles. The dataset covers 333 original examples of image-based wordplay, cluing 13 categories such as movies, composers, major cities, and food. To achieve good performance on the benchmark of identifying the clued word or phrase, models must combine image recognition and string manipulation with hypothesis testing, multi-step reasoning, and an understanding of human cognition, making for a complex, multimodal evaluation of capabilities. We find that GPT-4o significantly outperforms all other models, followed by proprietary models outperforming all other evaluated models. However, even the best model has a final accuracy of only 42\%, which goes down to just 7\% on hard puzzles, highlighting the need for substantial improvements in reasoning. Further, models rarely understand all parts of a puzzle, and are almost always incapable of retroactively explaining the correct answer. Our benchmark can therefore be used to identify major shortcomings in the knowledge and reasoning of multimodal large language models.
format Preprint
id arxiv_https___arxiv_org_abs_2401_05604
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle REBUS: A Robust Evaluation Benchmark of Understanding Symbols
Gritsevskiy, Andrew
Panickssery, Arjun
Kirtland, Aaron
Kauffman, Derik
Gundlach, Hans
Gritsevskaya, Irina
Cavanagh, Joe
Chiang, Jonathan
La Roux, Lydia
Hung, Michelle
Computation and Language
Artificial Intelligence
Computer Vision and Pattern Recognition
Computers and Society
We propose a new benchmark evaluating the performance of multimodal large language models on rebus puzzles. The dataset covers 333 original examples of image-based wordplay, cluing 13 categories such as movies, composers, major cities, and food. To achieve good performance on the benchmark of identifying the clued word or phrase, models must combine image recognition and string manipulation with hypothesis testing, multi-step reasoning, and an understanding of human cognition, making for a complex, multimodal evaluation of capabilities. We find that GPT-4o significantly outperforms all other models, followed by proprietary models outperforming all other evaluated models. However, even the best model has a final accuracy of only 42\%, which goes down to just 7\% on hard puzzles, highlighting the need for substantial improvements in reasoning. Further, models rarely understand all parts of a puzzle, and are almost always incapable of retroactively explaining the correct answer. Our benchmark can therefore be used to identify major shortcomings in the knowledge and reasoning of multimodal large language models.
title REBUS: A Robust Evaluation Benchmark of Understanding Symbols
topic Computation and Language
Artificial Intelligence
Computer Vision and Pattern Recognition
Computers and Society
url https://arxiv.org/abs/2401.05604