snap-stanford/humanual-email

---
license: cc-by-nc-4.0
language:
- en
tags:
- user-simulation
- humanlm
- persona
- business-communication
pretty_name: Humanual-Email
size_categories:
- 10K<n<100K
---

# Humanual-Email

[](https://humanlm.stanford.edu)
[](https://humanlm.stanford.edu/HumanLM_paper.pdf)
[](https://github.com/zou-group/humanlm)
[](https://huggingface.co/collections/snap-stanford/humanual-datasets)

Adapted from the Enron email dataset, capturing user communication in business settings including decision negotiation, project status reporting, and constraint resolution. This dataset is part of the **[HumanLM](https://humanlm.stanford.edu)** benchmark for training user simulators that accurately reflect real user behavior.

**Source:** Enron corpus · **Domain:** Business Communication · **Date Range:** 1974-01-04 to 2001-05-24

The dataset contains **7,043** comments from **399** users across **5,153** posts, with an average of **1.69** turns per conversation. Each example includes the user's persona, conversation context, and ground-truth response.

**Splits:** train (6,377) · val (130) · test (536)

| Column | Description |
|--------|-------------|
| `prompt` | Email thread context as a list of messages with `role` and `content` fields |
| `completion` | The ground-truth email reply to generate |
| `persona` | User's email communication style and role at Enron |
| `post_id` | Unique email thread identifier |
| `user_id` | SHA-256 hashed email address (for privacy) |
| `timestamp` | Unix timestamp of when the email was sent |
| `turn_id` | Position in the email thread (1 = first reply) |
| `metadata` | Email metadata as JSON (from, to, subject, cc, etc.) |

## Quick Start

```python
from datasets import load_dataset

dataset = load_dataset("snap-stanford/humanual-email")
sample = dataset["train"][0]

print(sample["persona"]) # User persona
print(sample["prompt"]) # Conversation context
print(sample["completion"]) # Ground-truth response
```

## Citation

```bibtex
@article{wu2026humanlm,
title={HUMANLM: Simulating Users with State Alignment Beats Response Imitation},
url={https://humanlm.stanford.edu/},
author={Wu, Shirley and Choi, Evelyn and Khatua, Arpandeep and Wang, Zhanghan and He-Yueya, Joy and Weerasooriya, Tharindu Cyril and Wei, Wei and Yang, Diyi and Leskovec, Jure and Zou, James},
year={2026}
}
```

Released under [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/).