This dataset contains 20 math problems selected from the MATH-500 dataset, focusing on problems with difficulty levels 1–3.
It includes the original solutions, final answers, and generated completions using various sampling methods along with their corresponding rewards and final answers.
Completions were generated using the model Qwen/Qwen2.5-1.5B-Instruct, and reward scoring was performed with the… See the full description on the dataset page: https://huggingface.co/datasets/samhouliston/Math20-WBON-Study.
