Dolma 1.5 Culture-Symbol Relevancy Classification
收藏Zenodo2024-12-10 更新2026-05-26 收录
下载链接：
https://zenodo.org/doi/10.5281/zenodo.14366489
下载链接
链接失效反馈官方服务：
资源简介：
This is supplementatry material for paper "Attributing Culture-Conditioned Generations to Pretraining Corpora", containing pretraining documents from Dolma 1.5 corpora that our MEMOed framework deem "relevant" to the memorization of a symbol for a culture. The data is formatted as:
{
"<symbol1>": {
"<culture1>": {
[<document_id>, <d-SNR score>, <d-SENT score>, <d-TOK score>],...
},
...
},
"<symbol2>": ...,
...
}
提供机构：
Zenodo
创建时间：
2024-12-10
