遇见数据集

long-proof-pile

收藏
魔搭社区2026-07-06 更新2026-08-02 收录
官方服务:

资源简介:

Documents from [Proof Pile](https://huggingface.co/datasets/EleutherAI/proof-pile-2) with over 128K tokens, as measured by the [starmie-v1](https://huggingface.co/moondream/starmie-v1) tokenizer. Intended use is to measure long context cross-entropy.

提供机构:
maas
创建时间:
2025-08-20
二维码
社区交流群
二维码
科研交流群
商业服务