AmazonScience/TISER

TISER数据集是一个用于训练语言模型进行时间推理的任务数据集。它通过一个多阶段的推理管道，结合显式推理、时间线构建和迭代自我反思，使模型能够在推理过程中适应并扩展其内部推理过程。数据集包含训练和测试数据，其中训练数据有详细的提示和输出，测试数据则只有答案用于评估模型的时间推理能力。

The TISER dataset is a task dataset for training language models to perform temporal reasoning. It utilizes a multi-stage inference pipeline that combines explicit reasoning, timeline construction, and iterative self-reflection to enable models to adapt and scale their internal reasoning process during inference. The dataset includes training and test data, with the training data providing detailed prompts and outputs, while the test data only includes answers for evaluating the models temporal reasoning capability.