TurtleBench: A Visual Programming Benchmark in Turtle Geometry
This addresses the need for robust evaluation methods for LMMs in visual and geometric reasoning, though it is incremental as it introduces a new benchmark rather than a model improvement.
The authors tackled the challenge of evaluating large multimodal models' ability to reason about geometric patterns by introducing TurtleBench, a benchmark based on turtle geometry, and found that leading models like GPT-4o achieved only 19% accuracy on simple tasks with minimal improvement from few-shot prompting.
Humans have the ability to reason about geometric patterns in images and scenes from a young age. However, developing large multimodal models (LMMs) capable of similar reasoning remains a challenge, highlighting the need for robust evaluation methods to assess these capabilities. We introduce \Turtle, a benchmark designed to evaluate LMMs' capacity to interpret geometric patterns -- given visual examples, textual instructions, or both -- and generate precise code outputs. Inspired by turtle geometry, a notion used to teach children foundational coding and geometric concepts, TurtleBench features tasks with patterned shapes that have underlying algorithmic logic. Our evaluation reveals that leading LMMs struggle significantly with these tasks, with GPT-4o achieving only 19\% accuracy on the simplest tasks and few-shot prompting only marginally improves their performance ($<2\%$). \Turtle highlights the gap between human and AI performance in intuitive and visual geometrical understanding, setting the stage for future research in this area. \Turtle stands as one of the few benchmarks to evaluate the integration of visual understanding and code generation capabilities in LMMs, setting the stage for future research. Code and Dataset for this paper is provided here: \href{https://github.com/sinaris76/TurtleBench}{https://github.com/sinaris76/TurtleBench}