mlfoundations-dev/multiple_samples_majority_consensus_numina_aime_llm_verifier

该数据集包含问题及其对应的解决方案，以及模型针对问题的响应。每个问题可能包含多个评分者和评分理由，同时记录了大多数评分者的响应。数据集还包含了验证过的模型响应和对话信息。数据集旨在用于训练和评估模型对问题的响应能力。

The dataset includes problems and their corresponding solutions, as well as the models responses to the problems. Each problem may have multiple evaluators and their reasoning for the scores, along with the majority evaluators responses recorded. The dataset also contains verified model responses and conversation information. It is intended for training and evaluating the models ability to respond to problems.