Where do machine learning researchers find preprints?

arXiv is the main preprint archive. Know Your ML shows a live arXiv feed, Hugging Face lists daily papers, and Semantic Scholar searches the scientific literature.

Which site ranks AI models?

Arena AI ranks models with human votes across text, code, image, video and vision. Epoch AI publishes benchmark results and trend data, which suits longer comparisons.

Is arXiv peer reviewed?

No. arXiv hosts preprints posted before or alongside review, so readers must check each paper. Know Your ML says its own reports are AI-written and not peer reviewed.

What is the Machine Learning Decision Toolkit?

A paid digital download from The Knowledge Project with scorecards, workload templates and deployment checks for choosing ML systems. It is a planning aid, not a research source.

Best Machine Learning Research Resources in 2026

The ranking

Top pickarXiv93.5
Best with a free planHugging Face90.0

1
arXiv
The primary archive for ML preprints, with PDFs, subject listings and bulk metadata for developers. Preprints are not peer reviewed, so readers must judge each paper themselves.
Best for Reading the newest machine learning preprints
WebFree
93.5of 100
Visit Alternatives
2
Hugging Face
A hub for models, datasets and papers, with a daily papers page and hosted inference. Quality varies by upload, and the PRO plan adds private storage and compute.
Best for Finding models, datasets and daily papers
WebFree plan, paid upgrades
90.0of 100
Visit Alternatives
3
Semantic Scholar
Free AI-assisted search across scientific literature, with a public API from Ai2. Some reader features are in beta, and it indexes papers rather than hosting preprints.
Best for Searching papers across scientific fields
WebFree
88.0of 100
Visit Alternatives
4
Epoch AI
Public datasets on AI models over time, with benchmarks and reports on trends. It is analysis rather than a live ranking, so pair it with Arena AI for current models.
Best for Long-run trends in AI capability and compute
WebFree
86.5of 100
Visit Alternatives
5
Know Your ML *
A reading room with a live arXiv feed, briefs and long-form reports that link to open-access papers. Its reports are AI-written and not peer reviewed, and the feed covers six arXiv categories.
Best for Scanning recent arXiv papers with filters
WebFree plan, paid upgrades
78.5of 100
Visit Alternatives
6
OpenReview
Open peer review platform that hosts submissions and reviews for a long list of machine learning and AI venues, with 171 venues on its index. Its review record is primary for those venues, but coverage depends on which conferences use it.
Best for Reviews and submissions for ML conferences
WebFree
85.5of 100
Visit Alternatives
7
Arena AI
Live human-vote leaderboards for text, code, image, video and vision models. Rankings shift as models are added, and they reflect voter preference.
Best for Comparing current models by human preference
WebFree
84.5of 100
Visit Alternatives
8
Google Scholar
Free search across articles, theses, books and court opinions, with author, journal and date filters. It is a broad index rather than a source of reviews, so readers must open each result, and its homepage shows no AI summaries.
Best for Broad search across scholarly literature
WebFree
83.5of 100
Visit Alternatives
9
Kaggle
Free datasets, competitions and notebooks with GPU access on Kaggle's own environment. It is a practical place to find data and test ideas, but community notebooks vary in quality, so it ranks below the primary research sources here.
Best for Datasets and free GPU notebooks for experiments
WebFree
82.0of 100
Visit Alternatives
10
Machine Learning Decision Toolkit *
Scorecards and templates for choosing ML systems on quality, latency, cost and risk. It is a paid download for one planning use, so it ranks below the free reading and data resources.
Best for Teams comparing ML options on trade-offs
WebPaid
66.5of 100
Visit Alternatives

At a glance

Product	Score	Price	Platforms	Main content	Developer API	Peer review
arXiv	93.5	Free	Web	Preprints	Yes, arXiv API	Not peer reviewed
Hugging Face	90.0	Free plan, paid upgrades	Web	Models, datasets and papers	Yes	Not stated
Semantic Scholar	88.0	Free	Web	Indexed papers	Yes, Academic Graph API	Not stated
Epoch AI	86.5	Free	Web	Datasets and reports	Not listed	Not stated
Know Your ML *	78.5	Free plan, paid upgrades	Web	Feed, briefs and reports	Not listed	Reports not peer reviewed
OpenReview	85.5	Free	Web	Submissions and peer reviews	Not listed	Peer review run on the platform
Arena AI	84.5	Free	Web	Model leaderboards	Not listed	Vote-based rankings
Google Scholar	83.5	Free	Web	Index of articles, theses and books	Not listed	Mixed; check each source
Kaggle	82.0	Free	Web	Datasets and notebooks	Not listed	Community-written, not peer reviewed
Machine Learning Decision Toolkit *	66.5	Paid	Web	Scorecards and templates	Not listed	Not applicable

How to choose machine learning research resources

Primary sources: Start with arXiv for preprints and Semantic Scholar for search; summaries and reports are secondary.
Leaderboards: Arena AI shows human-vote rankings and Epoch AI shows benchmark data and trends. Compare both before choosing a model.
Models and datasets: Hugging Face is the main place to find model weights and datasets, and each upload carries its own license.
Decision tools: Scorecards help a team compare ML options on cost, latency and risk, but they do not replace testing on your own workload.

What the scores weigh

Primary sources and checkability
Coverage of ML research
Free access
Use for choosing and evaluating models

Scores are editorial, from 0 to 100, and use these criteria. Full method · Disclosure

About machine learning research resources

Machine learning research moves through preprints first. arXiv is the main archive, and tools such as Know Your ML * and Semantic Scholar sit on top of it, adding feeds, briefs and search. Hugging Face hosts models, datasets and a daily papers page that many practitioners check.

Evaluation sources answer a different question: which models perform well, and how. Arena AI ranks models using human votes, while Epoch AI publishes datasets and benchmark results on longer trends. Rankings change as new models are added, so every ranking should be read with its date.

The ML Decision Toolkit is a paid digital download of scorecards and templates for choosing systems, rather than a research source. Scores here are editorial judgments against the criteria, not measurements of research quality.

Frequently asked questions

Where do machine learning researchers find preprints?: arXiv is the main preprint archive. Know Your ML * shows a live arXiv feed, Hugging Face lists daily papers, and Semantic Scholar searches the scientific literature.
Which site ranks AI models?: Arena AI ranks models with human votes across text, code, image, video and vision. Epoch AI publishes benchmark results and trend data, which suits longer comparisons.
Is arXiv peer reviewed?: No. arXiv hosts preprints posted before or alongside review, so readers must check each paper. Know Your ML * says its own reports are AI-written and not peer reviewed.
What is the Machine Learning Decision Toolkit?: A paid digital download from The Knowledge Project with scorecards, workload templates and deployment checks for choosing ML systems. It is a planning aid, not a research source.