Children's Intelligence Tests Pose Challenges for MLLMs? KidGym: A 2D Grid-Based Reasoning Benchmark for MLLMs

TL;DR

KidGym is a new 2D grid-based benchmark inspired by children's intelligence tests, designed to evaluate multiple core capabilities of Multimodal Large Language Models (MLLMs) in a comprehensive and customizable way.

Contribution

The paper introduces KidGym, a versatile benchmark with 12 tasks for assessing MLLMs' cognitive abilities, emphasizing adaptability and developmental potential.

Findings

State-of-the-art MLLMs show significant limitations on KidGym tasks.

KidGym's diverse scenarios reveal gaps in perception, reasoning, and memory capabilities.

Benchmark is fully customizable and extensible for future research.

Abstract

Multimodal Large Language Models (MLLMs) combine the linguistic strengths of LLMs with the ability to process multimodal data, enbaling them to address a broader range of visual tasks. Because MLLMs aim at more general, human-like competence than language-only models, we take inspiration from the Wechsler Intelligence Scales - an established battery for evaluating children by decomposing intelligence into interpretable, testable abilities. We introduce KidGym, a comprehensive 2D grid-based benchmark for assessing five essential capabilities of MLLMs: Execution, Perception Reasoning, Learning, Memory and Planning. The benchmark comprises 12 unique tasks, each targeting at least one core capability, specifically designed to guage MLLMs' adaptability and developmental potential, mirroring the stages of children's cognitive growth. Additionally, our tasks encompass diverse scenarios and objects with randomly generated layouts, ensuring a more accurate and robust evluation of MLLM capabilities. KidGym is designed to be fully user-customizable and extensible, allowing researchers to create new evaluation scenarios and adjust difficuly levels to accommodate the rapidly growing MLLM community. Through the evaluation of state-of-the-art MLLMs using KidGym, we identified significant insights into model capabilities and revealed several limitations of current models. We release our benchmark at: https://bobo-ye.github.io/KidGym/.