COLOSSEUM

← Benchmarks

Large-scale real-robot manipulation benchmark. Diverse tasks and environments.

Overview

COLOSSEUM is a real-robot benchmark with diverse manipulation tasks across multiple environments. Used to evaluate generalization and robustness of VLA and policy models. BridgeVLA achieves 64% success.

Related

• BridgeVLA — 64% on COLOSSEUM

Benchmarks compare policies in sim; our eval loop scores them on physical rigs.

RCSV Eval → License training data →