GeoChallenge: A Multi-Answer Multiple-Choice Benchmark for Geometric Reasoning with Diagrams

1. 1Generate multi-answer multiple-choice geometry proof problems using automated methods
2. 2Evaluate large language models on GeoChallenge to assess their geometric reasoning capabilities
3. 3Analyze results to identify areas of improvement for model development
4. 4Utilize GeoChallenge to develop more accurate and reliable geometric reasoning models

AI researchers and developers working on large language models can benefit from GeoChallenge to evaluate and improve their models' geometric reasoning capabilities, while data scientists can utilize this benchmark to develop more accurate models

💡 GeoChallenge provides a large-scale benchmark for evaluating geometric reasoning in LLMs, addressing the limitations of existing benchmarks

📐️ Introducing GeoChallenge: a benchmark for evaluating geometric reasoning in LLMs with 90K multi-answer multiple-choice questions and diagrams