Evidence categories
We label material claims by what supports them:
- Published research: a finding reported by a paper, dataset, or institutional study, with its scope and limitations.
- Engineering inference: a reasoned explanation derived from technical principles; it is not presented as an experimental result.
- SABHYA product design: a description of intended or implemented product behavior, with beta status and validation limits stated.
Source hierarchy
Prefer primary platform documentation, government and official sources, peer-reviewed research, and reputable institutional research. Use secondary reporting when a primary source is unavailable and label the source accordingly. Record publication and access dates for fast-changing platform details.
Claim review
Every quantitative claim requires a traceable source and a matching population, task, and evaluation method. A result from one dataset does not imply a production guarantee. We do not invent customer counts, testimonials, pricing, approval status, certifications, language coverage, or benchmark results.
Corrections and updates
Each research page should include a publication or review date, links to source material, and a contact route for corrections. When a source changes or a claim becomes outdated, update the page and record the change in the research backlog.