OVM-dataset

The training dataset for verifiers, which is generated by the finetuned models in GSM8K and Game of 24. The models are open-sourced in [OVM-llama2-7b](https://huggingface.co/FreedomIntelligence/OVM-llama2-7b) and [OVM-Mistral-7b](https://huggingface.co/FreedomIntelligence/OVM-Mistral-7b).

See the paper [Outcome-supervised Verifiers for Planning in Mathematical Reasoning](https://arxiv.org/pdf/2311.09724.pdf) and the code in [github](https://github.com/FreedomIntelligence/OVM)

本数据集为验证器训练数据集，由针对GSM8K与24点游戏（Game of 24）任务进行微调得到的模型生成。上述模型已分别在[OVM-llama2-7b](https://huggingface.co/FreedomIntelligence/OVM-llama2-7b)与[OVM-Mistral-7b](https://huggingface.co/FreedomIntelligence/OVM-Mistral-7b)中开源。相关研究细节可参阅论文《面向数学推理规划的结果监督验证器》（Outcome-supervised Verifiers for Planning in Mathematical Reasoning），以及代码仓库[GitHub](https://github.com/FreedomIntelligence/OVM)

搜集汇总

背景与挑战

背景概述

OVM-dataset是一个用于验证器的训练数据集，由在GSM8K和Game of 24任务上微调的模型生成，相关模型已开源。该数据集支持数学推理规划中的结果监督验证器研究，基于Apache 2.0许可证发布。