caskcsg/NExtLong-Instruct-dataset-Magpie-Llama-3.3-Pro-1M-v0.1
收藏Hugging Face2025-08-02 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/caskcsg/NExtLong-Instruct-dataset-Magpie-Llama-3.3-Pro-1M-v0.1
下载链接
链接失效反馈官方服务：
资源简介：
NExtLong数据集是由短文本数据集合成而成的，包括64K、128K和512K不同大小的合成数据集。这些数据集用于训练能够处理长上下文的模型，并通过硬负干扰项来增强模型对长距离依赖性的建模能力。
The NExtLong datasets are composed of synthetic data from short-text datasets, including datasets of sizes 64K, 128K, and 512K. These datasets are used to train models capable of handling long contexts, enhanced by the inclusion of hard negative distractors to improve the models ability to model long-range dependencies.
提供机构：
caskcsg
