Preview  Clio is being built in the open. Score aggregation is in curation. How we source it

Benchmarks

How the field measures itself — simulation suites, real-robot protocols, distributed arenas, and the pooled datasets underneath them. Start here to know what a reported number means.

Loading benchmarks…
Descriptors are drawn from each benchmark's own paper or site, linked on every row.

Reported results

Loading results…
Every number sits next to its source link and a provenance mark: official — the benchmark's own leaderboard; self-reported — the model's paper or release; reproduced — an independent re-run; third-party — measured by someone else. When sources disagree, every number is shown side by side — never averaged, never the best-looking one. How scores are admitted