priorcomputers/openreview_raw

---
dataset_info:
features:
- name: forum_id
dtype: string
- name: forum_title
dtype: string
- name: forum_authors
sequence: string
- name: forum_abstract
dtype: string
- name: forum_keywords
sequence: string
- name: forum_pdf_url
dtype: string
- name: forum_url
dtype: string
- name: note_id
dtype: string
- name: note_type
dtype: string
- name: note_created
dtype: int64
- name: note_replyto
dtype: string
- name: note_readers
sequence: string
- name: note_signatures
sequence: string
- name: venue
dtype: string
- name: year
dtype: string
- name: note_text
dtype: string
splits:
- name: train
num_bytes: 2565679898
num_examples: 626430
download_size: 758998779
dataset_size: 2565679898
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
license: odc-by
language:
- en
tags:
- peer-review
- openreview
- scientific-reviews
size_categories:
- 100K<n<1M
---

# OpenReview Raw

Raw peer review data from OpenReview, covering major ML/AI venues (ICLR, NeurIPS, EMNLP, COLM, ACM MM, and more). Includes reviews, official comments, meta-reviews, and decisions for 49,023 unique papers.

Originally from [`sumukshashidhar-archive/openreview_raw`](https://huggingface.co/datasets/sumukshashidhar-archive/openreview_raw).

This dataset is a compilation of publicly available data from OpenReview. All original content and data rights belong to OpenReview. This compilation is made available under the Open Data Commons Attribution License (ODC-By). Users must attribute both this compilation and the original source (OpenReview) in any use of this dataset.

## Dataset Statistics

| Statistic | Value |
|-----------|-------|
| Total rows | 626,430 |
| Unique papers | 49,023 |
| Unique venues | 349 |
| Year range | 2013–2025 |

### Note Types

| Type | Count | % |
|------|-------|---|
| official_comment | 349,653 | 55.8% |
| official_review | 186,462 | 29.8% |
| decision | 31,450 | 5.0% |
| review | 28,616 | 4.6% |
| comment | 16,753 | 2.7% |
| meta_review | 13,496 | 2.2% |

### Top Venues

| Venue | Count |
|-------|-------|
| ICLR 2025 | 198,960 |
| ICLR 2024 | 110,570 |
| NeurIPS 2024 | 75,555 |
| NeurIPS 2023 | 64,562 |
| EMNLP 2023 | 22,742 |
| NeurIPS 2022 | 16,278 |
| ICLR 2022 | 14,593 |
| NeurIPS 2021 | 13,605 |
| ICLR 2021 | 12,275 |
| ICLR 2019 | 11,916 |

### Year Distribution

| Year | Count |
|------|-------|
| 2013 | 373 |
| 2014 | 651 |
| 2016 | 295 |
| 2017 | 626 |
| 2018 | 1,158 |
| 2019 | 14,284 |
| 2020 | 12,979 |
| 2021 | 35,943 |
| 2022 | 44,621 |
| 2023 | 96,525 |
| 2024 | 219,635 |
| 2025 | 199,340 |

### Note Text Length (characters)

| Statistic | Value |
|-----------|-------|
| Mean | 2,268 |
| Median | 2,023 |
| Min | 10 |
| Max | 56,453 |

## Schema

- **forum_id** — OpenReview forum identifier (one per paper)
- **forum_title** — Paper title
- **forum_authors** — List of paper authors
- **forum_abstract** — Paper abstract
- **forum_keywords** — Paper keywords
- **forum_pdf_url** — Link to PDF on OpenReview
- **forum_url** — Link to forum on OpenReview
- **note_id** — Unique identifier for this note (review/comment/decision)
- **note_type** — One of: `official_review`, `official_comment`, `decision`, `review`, `comment`, `meta_review`
- **note_created** — Unix timestamp (milliseconds) of note creation
- **note_replyto** — ID of the note this is replying to
- **note_readers** — List of reader groups with access
- **note_signatures** — List of note author signatures
- **venue** — Conference/venue identifier
- **year** — Publication year
- **note_text** — Full text content of the note