Methodology
How RightForAI handles data, and why.
Traceable rankings
We preserve source rankings where available. When only task scores are provided, we rank results within the same task and snapshot. Each leaderboard identifies its source version.
Explicit configurations
Reasoning configurations such as effort and thinking are retained in source names rather than silently merged.
No combined rankings
Rankings from different evaluations, categories, and tasks are presented separately. We do not add or average them into a combined ranking.
Immutable, traceable snapshots
Raw data is saved as immutable, content-addressed snapshots. Processed results use source-specific model and record IDs to preserve version history.