Hwilner/imo-gradingbench
收藏官方服务：
资源简介：
IMO-GradingBench是一个评估大型语言模型自动评分能力的基准数据集，包含1000个数学问题的模型生成解决方案的人类评分。该数据集是IMO-Bench套件的一部分，由Google DeepMind发布。每个数据实例包括问题、解决方案、评分指南、人类评分者提供的分数和奖励。
IMO-GradingBench is a benchmark dataset for evaluating the automatic grading capabilities of large language models, containing 1,000 human gradings of model-generated solutions to mathematical problems. It is part of the IMO-Bench suite released by Google DeepMind. Each data instance includes a problem, a solution, grading guidelines, and points and reward provided by human graders.
提供机构：
Hwilner
