SYNTOLOGY The Modern Ontology for AI HomeExplorerDemoMethodologyEthosAboutDevelopers

Search the graph, then open only what you want to check.

Results start closed. Every row expands to the evidence behind it — the edges that were written, where each one came from, the paper's own sentence, and the review record as reviewers left it. Nothing here is a derived summary you cannot check. This is a fixed six-paper slice showing the shape of the output — a static walkthrough of the UI, not a live query. The graph itself is real and live now (see the live counts to the right); query it directly for a live answer via the API or MCP — see /developers.

25,045
papers parsed into the schema
99.47%
structural success, verified end to end
134
failures out of 25,179, individually diagnosed
{{ matchLabel }}
Venue
Trace readiness
Evidence present
{{ p.num }} {{ p.title }} {{ p.venue }} {{ p.year }} {{ p.readiness }} {{ p.tipText }} {{ p.caret }}
Edges written · each one sourced
{{ e.rel }} {{ e.prov }}
Implementation
{{ p.repo }}
Captured from the paper's own text at ingestion, before popularity signals accumulated. Maturity: {{ p.codeAge }}.
Citation context — the paper's own sentence

{{ p.quote }}

{{ p.quoteFrom }}
Review record — never averaged
{{ r.rating }}
conf {{ r.confidence }}
Kept as a distribution. Averaging across reviewer confidence is a known, biased statistic.

Indexing covers NeurIPS, ICML, ICLR, CVPR, ICCV, ECCV, ACL, EMNLP, NAACL and arXiv. The slice shown here is drawn from the general-ML core, which is the part measured end to end. Coverage is broad but not comprehensive, and is not presented as such.
Self-supervised visual representation · 7 nodes hover a node to isolate its edges dashed = model-derived edge, confidence recorded
{{ n.label }}
{{ n.meta }}
{{ l.text }}
Days from publication to the graph, not months.
Read the methodology
Identity isn't tied to activityQueries counted in aggregate, then discardedRanking and labeling methods disclosedmedia@syntology.aisyntology.ai