Paper Rankings.across live arXiv categories.
Search and explore AI-assessed scientific preprints. Every paper is rated on 16 dimensions covering impact, quality and other useful criteria.
Validation Report by Swiss Economics (ICLR 2026 correlation)Top Papers of the Week
| # | Paper | |||||||
|---|---|---|---|---|---|---|---|---|
| 1 | Super-Brownian limits and the -point function for high-dimensional percolation Arthur Blanc-Renaudie, Tom Hutchcroft | 8.5 | 9.0 | 8.5 | 4.0 | 1.0 | 7.0 | |
| 2 | Efficient classical simulation of large-scale unitary cluster Jastrow circuits Hrishikesh Belagali, Thomas Van Camp +4 | 8.0 | 8.0 | 7.0 | 7.5 | 6.0 | 6.0 | |
| 3 | On Feige's conjecture Zipei Nie, Jiaye Wei | 8.0 | 8.0 | 8.5 | 6.0 | 2.5 | 7.0 | |
| 4 | Unexpected Collisional Rotational Excitation via Long-Range Capture and Orbiting Dasarath Swaraj, Guodong Zhang +6 | 8.0 | 8.5 | 8.0 | 8.0 | 4.0 | 5.5 | |
| 5 | Observable Estimation in the Absence of Classical Verification Samantha V. Barron, Bradley Mitchell +6 | 8.0 | 8.0 | 7.5 | 5.0 | 7.0 | 6.5 | |
| 6 | Elemental Germanium Phase-Change Memory Till Zellweger, Marko Mladenović +6 | 8.0 | 8.0 | 9.0 | 8.5 | 8.0 | 6.5 | |
| 7 | Bipartite Bound Information Exists Jef Pauwels, Nicolas Gisin +1 | 8.0 | 8.5 | 9.0 | 7.0 | 2.0 | 6.5 | |
| 8 | A generalized Kirchhoff's law of thermal radiation for Floquet media Sander A. Mann, Dimitrios L. Sounas +1 | 8.0 | 8.5 | 8.5 | 7.0 | 4.0 | 8.0 |
Explore newly ranked papers and active research categories.
Newly Ranked Papers
Latest papers added to the live Kurate ranking pipeline.
+40 more categories
20,980 papers across all remaining fields.
What Kurate supports today.
A ranking and discovery layer for scientific preprints. The capabilities below reflect what is currently live on the platform.
16 metrics for every paper
Papers are rated on 16 dimensions covering quality, impact, and practical relevance: significance, rigor, novelty, clarity, reproducibility, evidence strength and more. Each score comes with a short written justification.
Frontier AI assessment
Claude Opus 4.8 reads the full paper text and rates each dimension independently, with a written justification behind every score.
Interactive heatmap explorer
Tens of thousands of papers in a color-coded grid. Sort by any metric, set thresholds, filter by category and date, or search titles and authors. Everything updates instantly.
Impact rating (1 to 10)
A single predicted impact rating summarises each paper. The finer-grained dimensions behind it show where that impact comes from.
Per-metric filtering & thresholds
Combine filters to ask precise questions, like highly novel and rigorous papers in your field, or work that challenges earlier findings.
Live statistics & correlations
Distributions and correlations across the whole dataset, refreshed automatically as new papers are assessed.
Planned features under active development. These are not yet part of the live platform.
- Metric-based category leaderboards
- Field-level validation reporting
- Cross-model agreement metrics
- Customizable email digests
A discovery workflow for fast-moving scientific literature.
Kurate assesses every paper on 16 dimensions using frontier AI models and presents the results in an explorable heatmap, so researchers can quickly spot work worth a closer look.
Collect papers
Kurate gathers scientific preprints from supported arXiv categories as new work is published.
AI reads each paper
Claude Opus 4.8 reads the full paper text and assesses it on 16 dimensions, from significance and methodological rigor to how surprising, reproducible, and foundational the work is.
Score every dimension
Each dimension gets a score from 1 to 10 with a short justification, or null where it does not apply. An overall impact rating summarises the paper.
Explore the heatmap
Browse every assessed paper in a color-coded heatmap. Sort by any metric, set thresholds, filter by category or date, and check live distributions and correlations.
Scientific papers are difficult to compare across unrelated fields.
A robotics paper, a quantum physics paper, and an economics paper may all be important, but they should not be interpreted through the same field assumptions. Kurate uses category-based leaderboards so papers are ranked within more meaningful research contexts.
Field context matters
A paper's significance is easier to interpret when compared with other papers from the same arXiv category.
Broad rankings hide specialised work
Important papers in smaller technical fields may be missed when discovery depends only on general popularity or social attention.
Category filters improve discovery
Move directly into the arXiv category you care about and inspect ranked papers within that context.
A ranking layer for scientific preprints.
Kurate adds a ranking layer on top of preprint discovery, combining category-based heatmaps with 16-dimension AI assessment so users can explore work that may deserve closer reading.
Built for researchers, students, supervisors, labs, institutions, and analysts.
Researchers
Follow fast-moving fields, identify ranked preprints, and discover papers that may not yet have citation visibility.
Postgraduate students
Scan active arXiv categories, find relevant preprints for literature discovery, and follow which topics are moving quickly.
Research supervisors
Recommend recent ranked papers, monitor category activity, and identify emerging work for discussion.
Research groups & labs
Track category activity, compare ranked papers within a field, and support reading-group paper selection.
Institutions
Monitor emerging scientific areas and where attention is forming across research categories.
Science communicators
Identify ranked papers that may become important and follow early signals in the scientific literature.
A discovery layer, not a replacement for peer review.
Kurate rankings are discovery signals, not peer review. They help you decide which papers deserve a closer read. You should still read the paper and judge the methodology, evidence, and limitations yourself.
Methodology, scope, and how to interpret Kurate signals.
Practical answers about what the platform does, how rankings are produced, and how they should be used in research workflows.