Datasets:
Tasks:
Text Retrieval
Modalities:
Text
Formats:
parquet
Languages:
English
Size:
10K - 100K
License:
File size: 1,671 Bytes
43791b5 c6ce56c 43791b5 c6ce56c 43791b5 c6ce56c 43791b5 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 | ---
license: cc-by-4.0
task_categories:
- text-retrieval
language:
- en
tags:
- citation
- retrieval
- arxiv
- neurips
- benchmark
pretty_name: FORAGE
configs:
- config_name: neurips25
data_files:
- split: cited_papers
path: neurips25/cited_papers.parquet
- split: source_papers
path: neurips25/source_papers.parquet
---
# The FORAGE Dataset
FORAGE (Finding Out Related Articles via Generative Exploration) is an agentic retrieval benchmark based on the related works prediction task.
## Subset: `neurips25`
| Split | Rows | Description |
|---|---|---|
| `source_papers` | 1,024 | NeurIPS 2025 papers used as queries |
| `cited_papers` | 25,050 | Retrieval corpus of cited papers |
### `source_papers` columns
| Column | Type | Description |
|---|---|---|
| `id` | string | NeurIPS 2025 submission ID |
| `content` | string | Full paper text (Markdown) |
| `arxiv_id` | string | arXiv identifier |
| `title` | string | Paper title |
| `abstract` | string | Paper abstract |
| `author` | string | Semicolon-separated author names |
| `first_author_last_name` | string | Last name of first author |
| `related_work` | list[string] | arXiv IDs cited in Related Work — **primary eval target** |
| `all_cited` | list[string] | arXiv IDs cited anywhere in the paper |
### `cited_papers` columns
| Column | Type | Description |
|---|---|---|
| `id` | string | arXiv paper ID |
| `content` | string | Full paper text (Markdown) |
| `title` | string | Paper title |
| `abstract` | string | Paper abstract |
| `first_author_last_name` | string | Last name of first author |
| `year` | int32 | Publication year |
| `created` | date | arXiv submission date |
|