The Karp Dataset
This provides a new benchmark for assessing and enhancing reasoning capabilities in LLMs, though it is incremental as it focuses on a specific domain of mathematical proofs.
The authors tackled the need for datasets to evaluate mathematical reasoning in LLMs by introducing the Karp dataset, which consists of detailed NP-completeness reduction proofs, and they showed that fine-tuning with it improves model performance.
Understanding the mathematical reasoning capabilities of Large Language Models (LLMs) is a central topic in the study of artificial intelligence. This new domain necessitates the creation of datasets of reasoning tasks for both training and benchmarking the performance of LLMs. To this end, we introduce the Karp dataset: The first dataset composed of detailed proofs of NP-completeness reductions. The reductions vary in difficulty, ranging from simple exercises of undergraduate courses to more challenging reductions from academic papers. We compare the performance of state-of-the-art models on this task and demonstrate the effect of fine-tuning with the Karp dataset on reasoning capacity.