kyutai/ARC_finetuning
收藏Hugging Face2025-10-24 更新2025-10-18 收录
下载链接：
https://hf-mirror.com/datasets/kyutai/ARC_finetuning
下载链接
链接失效反馈官方服务：
资源简介：
ARC-Encoder微调数据集包含12个jsonl文件，分为翻译、问题回答、阅读理解和总结四个任务类别。该数据集用于微调ARC-Encoder模型，以学习压缩文本表示，支持大型语言模型。数据集汇集了多个现有数据集的样本，包括对抗性QA、FreebaseQA、ASQA等，适用于独立地从每个数据集收集上下文示例，设计最终的微调样本。
The ARC-Encoder finetuning dataset consists of 12 jsonl files across four task categories: Translation, Question-Answering, Reading Comprehension, and Summarization. It is used to fine-tune the ARC-Encoder model for learning compressed text representations that support large language models. The dataset aggregates samples from various existing datasets such as AdversarialQA, FreebaseQA, ASQA, etc., allowing for independent collection of in-context examples from each dataset to design the final finetuning samples.
提供机构：
kyutai
