Update README.md

README.md

@@ -15,14 +15,13 @@ dataset_info:
   dataset_size: 76605258
 ---
 
-# ArXiv papers from The Pile for document-level membership inference for LLMs (split into sequences)
+# ArXiv papers from The Pile for document-level MIAs against for LLMs (split into sequences)
 
-This dataset contains **sequences** from ArXiv papers randomly sampled from the train (members) and test (non-members) dataset from (the uncopyrighted version of) [the Pile](https://huggingface.co/datasets/monology/pile-uncopyrighted). 
-We randomly sample 1,000 documents from the train set (members) and 1,000 documents from the test set (non-members), ensuring that the selected documents have at least 5,000 words (any sequences of characters seperated by a white space). 
-In this dataset, we split each document in 25 sequences of 200 words. 
+This dataset contains **sequences from** ArXiv papers randomly sampled from the train (members) and test (non-members) dataset from (the uncopyrighted version of) [the Pile](https://huggingface.co/datasets/monology/pile-uncopyrighted). 
+We randomly sample 1,000 documents members and 1,000 non-members, ensuring that the selected documents have at least 5,000 words (any sequences of characters seperated by a white space). 
+This dataset contains the first 25 sequences of 200 words from all the documents made available in full [here](https://huggingface.co/datasets/imperial-cpg/pile_arxiv_doc_mia).
 
-As such, the dataset can be used to develop and evaluate document-level MIAs against LLMs trained on The Pile. 
+The dataset can be used to develop and evaluate document-level MIAs against LLMs trained on The Pile. 
+Target models include the suite of Pythia and GPTNeo models, to be found [here](https://huggingface.co/EleutherAI). Our understanding is that the deduplication executed on the Pile to create the "Pythia-dedup" models has been only done on the training dataset, suggesting this dataset of members/non-members also to be valid for these models. 
 
-Target models include the suite of Pythia and GPTNeo models, to be found [here](https://huggingface.co/EleutherAI). 
-
-Note: our understanding is that the deduplication executed on the Pile to create the "Pythia-dedup" models has been only done on the training dataset, suggesting this dataset of members/non-members also to be valid for these models. 
+For more information we refer to [the paper](https://arxiv.org/pdf/2406.17975).