Live on HuggingFace · datasets shipped
Describe a research area in plain English. FineSet pulls from arXiv, Semantic Scholar, GitHub and more — merges duplicates, scores them, and hands you clean JSONL or Parquet, refreshed on a schedule.
Pulls fromarXivSemantic ScholarGitHub+ othersShipsJSONLParquetHF push
4datasets live on HF
4,800+papers assembled
weeklyrefresh cadence
0.92dedup similarity cutoff
The problem
You know the data exists. Getting it clean, deduplicated, and fresh is a week of glue code you rewrite for every project.
Doing it yourself
With FineSet
The pipeline
A topic you type becomes a clean dataset in three steps. No scrapers to write, nothing to maintain.
01
Give it keywords and, if you want, arXiv categories. Something like "RLHF, cs.LG, since 2023".
02
It pulls from every source that fits, normalizes the fields, merges duplicates, scores quality, and strips PII.
03
Export it and it stays fresh, refreshed every day. Parquet and a one-click push to HuggingFace are coming soon.
Try it
No signup. Animals, cars, LLMs, RLHF — type anything. We check the shared pool first, and if it's a topic we don't cover yet we pull a live sample across arXiv, GitHub and Hacker News — papers, repos and threads.
What's inside
The boring, necessary work — done before the file reaches you, every refresh.
Subscribe to a topic and new matching records flow in on a schedule. The dataset is never the stale copy you downloaded last month.
The same record from different sources merges into one row, with every source kept — pgvector similarity at a 0.92 cutoff.
Every record carries a 0–1 score from citation signal, so you can filter the noise out before you ever start training.
One clean record per line, normalized fields, ready for your loader. Parquet and one-click push to HuggingFace are next.
Emails, phone numbers, and other PII are removed before anything is stored — so what you train on stays clean.
Not just our templates. Describe any research area or data domain and FineSet builds the pipeline for it.
Live now
Download a sample in the exact JSONL you'd export — no account. The full dataset is a free account away.
Who it's for
Same pipeline, different output. If you need domain-specific records and you don't want to babysit scrapers, it's for you.
Thousands of papers on your topic as a JSONL dataset that refreshes on a schedule. Quality scored, so you can filter out the noise before you train.
e.g. “RLHF, cs.LG, since 2023”
Every new paper in your subfield, structured and deduplicated, waiting for you in the morning.
e.g. “mechanistic interpretability”
Tell it what data you want to track and the records keep flowing. No scrapers to write, none to babysit.
e.g. “LLM agents + tool use”
Pricing
No plans, no limits to weigh up. Build datasets and export them at no cost while we're early. Paid tiers come later — you'll hear about them first.
Free while in beta
$0
Every dataset, every export — free for now. No card, no plans to pick. We'll give plenty of notice before anything costs money.
FAQ
Pipelines refresh on a schedule. New papers land in your dataset the week they post to arXiv — no scraper to maintain.
JSONL today, with Parquet coming soon. The shape matches what's already on our HuggingFace datasets, so it's drop-in.
FineSet aggregates open metadata (arXiv, Semantic Scholar) and records each source. You're responsible for honoring each source's license for your use.
Create a pipeline for it. We configure the sources, run the first refresh, and keep it updated — the dataset builds itself from there.
Deterministic and transparent: derived from citation signal (log-scaled), not an opaque model. You can filter on it at export.
Yes — the same endpoints the dashboard uses, so you can wire pipelines and exports into your own workflow.
Subscribe to the shared pool, export your first JSONL, and spin up your own pipeline when you're ready. No credit card required.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。