ELI RESEARCH
Eli Research methodology and data principles
If we did not measure it, we do not publish it as a finding.
Research governance is active. Findings require captured observations and reproducible rules before publication.
RELEASE STANDARD
Methodology is publishable before findings are.
Eli can publish the question universe, schema, definitions and quality controls before a benchmark release. It will only publish an aggregate when the underlying observations are complete, reviewable and described with their coverage and limitations. Until then, this page makes no performance, ranking or customer-outcome claim.
Observations and definitions
Save evidence first, classify second, aggregate third and interpret last. Define recommendation, mention, citation, source type, action readiness and evidence tier without mid-report changes.
Version and uncertainty
Version prompt universe, methodology, dataset, report and scoring. State sample, missingness, ambiguous cases, model changes and limitations.
No fabricated benchmarks
Never present illustrative percentages, fake customer outcomes, invented rankings or synthetic rows as findings. Mark synthetic examples Illustrative example.
Corrections and sensitive data
Document errors, release corrected version and preserve history. Customer/private data needs permission, minimization and appropriate aggregation; security findings require responsible handling.
FAQ
Questions about the research design and release boundary.
AI research?
AI can assist, but findings require captured data/method.
Raw data always public?
Where appropriate.
Errors?
Versioned corrections.
Customers influence?
Methods prevent silent custom prompt changes.
Why limits?
Benchmarks need boundaries.
CONNECT YOUR WEBSITE
Explore Eli Research
Eli Research uses versioned methods, explicit sample definitions, inspectable raw evidence, transparent formulas, uncertainty, corrections and a strict separation between measured findings and hypotheses.
Explore Eli Research