The day-by-day upvote history of every Hugging Face Daily Papers entry since 2024-03-12, updated every day. It powers Paper Pulse (Space).
Dataset Card
By Tardelli Stekel, published under cc-by-4.0, revision f8f9c0d2bc0d.
The day-by-day upvote history of every Hugging Face Daily Papers entry since 2024-03-12, updated every day. It powers Paper Pulse (Space).
Built from the past revisions of hysts-bot-data/daily-papers-stats and the paper list in hysts-bot-data/daily-papers, both by @hysts, plus paper details from the public Daily Papers API. Thanks to hysts for keeping that record and for releasing it under CC0 1.0.
Files
| File | Rows | What it holds |
|---|---|---|
series.parquet |
one per paper per day | id (arXiv id), day, upvotes, comments: the count shown on the paper's page that day (last snapshot of the day, UTC) |
papers.parquet |
one per paper | title, Daily Papers date, authors (with Hugging Face accounts when linked), organization, submitter, GitHub repo and stars, abstract (summary), upvotes on day 0/1/3/7, up_close (7 days after the Daily Papers date), peak, purged, and rank within its day, ISO week and month |
purges.parquet |
one per event | days when the Hub removed votes from many papers at once: papers affected (n_drop), votes removed, and whether it was deep |
hourly.parquet |
one per paper per hour | id, ts (UTC), upvotes for the papers featured in the last 10 days, from the hourly snapshots of daily-papers-stats |
age_stats.json |
how papers usually do by age: the distribution of upvotes at the end of day 0..7 after their Daily Papers date, and what papers with a given count at day k ended their first week with (ratios by count bin) | |
meta.json |
last day, number of papers, purge days |
Notes
- Upvotes can go down. People unvote, and on a few days the Hub removed votes from many papers at once (2025-09-12, 2025-12-23, 2026-09-16, 2026-09-18; the cause was not announced). The series keeps the raw value of each day and never corrects it. A day counts as a purge when at least max(250, 2% of papers) dropped.
- Rankings use
up_close, the upvotes 7 days after the Daily Papers date, so a later purge does not rewrite past rankings. Papers younger than 7 days are ranked by their current count, and so are the papers whose Daily Papers date is before the history starts (March 12, 2024), which are flagged withearly. - The source has two collection gaps (13 and 7 days). Days without a snapshot have no row.
- Only papers that were featured on Daily Papers are included (about 18,500), not every paper on the Hub.
License
The series, rankings and purge records are released under CC BY 4.0. The source datasets by @hysts are CC0 1.0. Paper titles and abstracts (title, summary) belong to their authors and are included only to identify each paper; they are not covered by this dataset's license.
Structure
papers 19,253 rows
| Split | Rows | Size |
|---|---|---|
| train | 19,253 | 45.1 MB |
purges 4 rows
| Split | Rows | Size |
|---|---|---|
| train | 4 | 390 B |
series 8,083,260 rows
| Split | Rows | Size |
|---|---|---|
| train | 8,083,260 | 10.8 MB |
Details
- Repository
- tardellirs/paper-pulse-data
- Publisher
- Tardelli Stekel
- Task category
- Not stated by the source
- Tags
- papers, daily-papers, arxiv
- Size category
- 1M<n<10M
- Languages
- Not stated by the source
- Revision
- f8f9c0d2bc0d1b8b716e62bdff377b6cd3b1b910
- Last updated
- 2026-10-17
Files
8 files, 15.8 MB in total.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| age_stats.json | Data | 19.7 KB | — |
| hourly.parquet | Data | 22.2 KB | a2511ca72560 |
| meta.json | Data | 196 B | — |
| papers.parquet | Data | 15.0 MB | 0072fa00c55a |
| purges.parquet | Data | 2.1 KB | 2a6ffed62c72 |
| series.parquet | Data | 749.3 KB | 15a377bca347 |
| README.md | Documentation | 3.4 KB | — |
| .gitattributes | Repository | 2.5 KB | — |
License and Download
- License
- cc-by-4.0
- Access
- No access gate
Released by Tardelli Stekel through its official repository on Hugging Face. Read the license.