HINT-lab/Llama_3.1-8B-Instruct-Self-Calibration

本数据集包含了多种配置和特性，用于研究项目。每个数据集都有特定的特征、数据分割和文件路径。项目还涉及各种用于数据创建、模型训练和评估的脚本和配置。研究集中在一种名为Self-Calibration的有效测试时间缩放方法，该方法利用模型置信度进行动态采样调整。

The dataset includes multiple configurations and features for the research project. Each dataset has specific features, data splits, and file paths. The project also involves various scripts and configurations for data creation, model training, and evaluation. The research focuses on an efficient test-time scaling method called Self-Calibration, which utilizes model confidence for dynamic sampling adjustment.