RelBench Leaderboard

The RelBench leaderboard (submit ↗) uses self-reported prediction files. We place no restriction on submissions (e.g. runtime, hyperparameter-tuning, etc.) except the following:

1. For all submissions, any info at or in the future of the prediction time is off-limits.
2. For in-context submissions, gradient-based training on the target DB is not allowed.

For a more controlled and transparent evaluation on RelBench tasks, please see RelArena ↗.