datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
enron_spamThis is a version of the Enron Spam Email Dataset, containing emails (subject + message) and a label whether it is spam or ham.
enron-qa-emails-dasovich-jenron-qa-questions-dasovich-jenron-emailThis dataset includes emails from Enron Email Dataset with prompts processed from Are Large Pre-Trained Language Models Leaking Your Personal Information?.
To use the dataset, you can run the following in LLM-PBE.
from data.enron import EnronDataset
ds = EnronDataset(data_path="data/enron", pseudonymize=False)
enronhop-singlehopemail-Enron
email-Enron
Zenodo | Cornell | Source Paper
email-Enron is an undirected hypergraph built from the Enron email corpus, designed for higher-order network / hypergraph machine learning. In email communication, a single message can involve more than two people; this dataset captures that group interaction by modeling each email as a hyperedge connecting the sender and all recipients, while nodes represent Enron email addresses (restricted to a core set of employees).
Usage… See the full description on the dataset page: https://huggingface.co/datasets/anonymous-author-1234/email-Enron.email-Enron
email-Enron
Zenodo | Cornell | Source Paper
email-Enron is an undirected hypergraph built from the Enron email corpus, designed for higher-order network / hypergraph machine learning. In email communication, a single message can involve more than two people; this dataset captures that group interaction by modeling each email as a hyperedge connecting the sender and all recipients, while nodes represent Enron email addresses (restricted to a core set of employees).
Usage… See the full description on the dataset page: https://huggingface.co/datasets/daqh/email-Enron.hard-questions-enronqa
Hard Questions EnronQA
This dataset contains 138 questions where a cross encoder was able to achieve recall @ 5, but not recall @ 1!
The candidate documents have further been preprocssed with a summarization inference that summarizes the relevance of the candidate document with respect to the query. This is done to reduce the input length of the emails and also improves recall @ 1 significantly out of the box.
Sampled from EnronQA, Ryan et al. 2025.
pile_enronenronhop-timesplit_2001_0418
EnronHop — Time Split (cutoff: 2001-04-18)
Time-based re-split of the miluki/enronhop multi-hop Q&A benchmark, enabling temporal generalization evaluation.
Split policy
Items from the combined miluki/enronhop train+test pool (1,947 items total) are re-partitioned at cutoff 2001-04-18 00:00:00:
train (1,360 items, 69.85%): max(evidence_dates) < cutoff
test (587 items, 30.15%): max(evidence_dates) >= cutoff
evidence_dates are looked up per item by joining evidence_mails… See the full description on the dataset page: https://huggingface.co/datasets/DeL-TaiseiOzaki/enronhop-timesplit_2001_0418.enronhopenronhop-corpusenron_spamThis is a version of the Enron Spam Email Dataset, containing emails (subject + message) and a label whether it is spam or ham.
enronhop-corpusenronhopenron_dev_tstenron-cleaned