Article Quality Evaluator - Stabilarity Hub

World’s most detailed open paper quality evaluator

Paste any research paper URL — arXiv, journal DOI, PDF, PubMed — and get a full quality report in seconds. Free, no login, no limits.

📊 24 Quality BadgesReference quality, content structure, freshness, open access, code availability, and more — scored against the STABIL framework.

🔗 Reference GraphInteractive knowledge graph of all citations — nodes sized by trust score, edges showing citation relationships, click for full ref details.

🔍 Similarity CheckTF-IDF cosine comparison against 350+ indexed articles. Detects topic overlap and derivative work patterns.

🤖 LLM AnalysisOptional Haiku-powered consistency checks — claim density and evidence ratio as structured measurable outputs.

Supports: arXiv, DOI links, IEEE, ACM, Springer, Nature, MDPI, PubMed, direct PDF URLs, and most HTML pages.

Evaluate a Paper

Try a landmark paper:

Attention Is All You Need GPT-3 GPT-4 Report BERT LLaMA Mistral 7B

Fetching…

—

Unified Quality Score

trust×60 + structure×30 + optional×10

—

Reference Trust Score

weighted: t×30 a×25 i×20 l×15 s×5 b×5

—

Required Criteria

structural requirements passed

Badge Assessment

■ pass ■ fail ■ partial ■ n/a dashed = LLM / limited (manual only) · hover for details

Reference Analysis

Content Structure

Similarity to Known Articles

Reference Knowledge Graph

Force-directed graph of all citations. Click any node to see full details in the side panel. Drag to pin position. Scroll to zoom. Shift+drag to pan. Connections: grey = direct cite · blue = same domain · green = same journal.

Run evaluation to see knowledge graph

How to read this report:
A Unified Score of 60+ indicates a well-structured paper with quality references. arXiv preprints typically score 55–68 (no formal peer-review record). Journal papers with DOIs, open access, and diverse citations score 70+. The Trust Score reflects your reference list quality — papers citing peer-reviewed journals with DOIs score highest. LLM badges (dashed border) are optional manual checks — they don’t affect the auto-computed score.

Badge Reference Guide

Badge	Name	Type	What it measures	Threshold
REFERENCE QUALITY BADGES
`[s]`	Peer-Reviewed Sources	ref+range	% of refs from editorially peer-reviewed venues	≥80%
`[t]`	Trusted Sources	ref+range	% of refs from verified high-quality academic sources	≥80%
`[a]`	DOI Coverage	ref+range	% of refs with permanent Digital Object Identifiers	≥80%
`[b]`	CrossRef Indexed	ref+range	% of refs indexed in CrossRef scholarly database	≥80%
`[i]`	Metadata Indexed	ref+range	% of refs with complete bibliographic metadata	≥80%
`[l]`	Academic Venues	ref+range	% of refs from journals, conferences, or preprint servers	≥80%
`[f]`	Open Access	ref+range	% of refs freely accessible to all readers	≥80%
`[r]`	Reference Count	ref+range	Minimum number of references found	≥10
`[dn]`	Domain Diversity	ref+range	Unique source domains among references (avoids single-source bias)	≥5 domains
`[ln]`	Link Health	ref+range	% of reference URLs returning HTTP 200 (not broken)	≥80%
CONTENT & STRUCTURE BADGES
`[w]`	Word Count	article+range	Article length — substantive research requires depth	≥2,000
`[d]`	Registered ID	article+bool	DOI, arXiv ID, or other permanent identifier present	Present
`[o]`	Author ORCID	article+bool	Author ORCID iD found in article	Present
`[h]`	Reference Freshness	article+range	≥60% of refs from the last 2 years (relative to evaluation date)	≥60%
`[c]`	Data Visualizations	article+range	Original charts, figures, or tables (min. 2)	≥2
`[g]`	Code Repository	article+bool	Link to reproducible code (GitHub, GitLab, etc.)	Present
`[m]`	Diagrams	article+bool	Architecture or flow diagrams (Mermaid or image-based)	Present
`[ab]`	Abstract	article+bool	Structured abstract section detected	Present
`[p]`	Peer Reviewed	article+bool	Published in a known peer-reviewed venue (inferred from URL/DOI)	Yes
QUALITY CHECKS
`[ap]`	Anti-Plagiarism	quality+range	Cosine similarity vs. known articles — lower is better	<70% sim.
`[st]`	Structure Score	quality+range	Presence of standard academic sections (Abstract, Intro, Methods…)	≥3/6
`[ci]`	Citation Balance	quality+range	Self-citation ratio — single-domain over-reliance flagged	<20%
`[rc]`	LLM Consistency	quality+llm limited	Haiku checks claim-reference consistency (manual trigger, structured output)	Manual

Release Notes

v2.1 — 2026-04-07

Added 3 new deterministic badges: [ab] abstract detection, [dn] domain diversity, [ln] link health
All 24 badges now tagged with type (ref+range, article+bool, quality+llm+limited, etc.)
[rc] LLM consistency badge: marked as limited — excluded from auto-score, manual trigger only
[p] peer-review badge: now infers from publication venue URL (not text search) — arXiv = preprint, IEEE/Springer/Nature = peer-reviewed
LLM badges carry zero weight in Unified Score computation; limited badges excluded from req/opt counts
Required criteria expanded to 7 (added [ab])

v2.0 — 2026-04-06

Full arXiv HTML support — fetches complete paper text (5,000+ words) not just abstract
Bibliography parser: extracts arXiv IDs and DOIs from ltx_biblist structure
Domain-based ref scoring heuristics for refs not in DB
SEO meta descriptions, ScholarlyArticle schema, sitemap integration
Preprint submission pipeline (OSF + arXiv package generator)

v1.0 — 2026-04-05

Initial release: STABIL badge scoring, TF-IDF similarity, PDF support, auto-DB sync