RoboColosseum details: 5 VLAs ranked across 20,000+ real-world runs

DJiafei · x · 2026-11-06

Official thread on RoboColosseum, the open robot eval initiative by Hugging Face, BitRobot Network, EAIC and others: 100+ robots, 140+ in-the-wild scenes, open rollout data and 10,000 free eval hours. Two tracks—blind open-world zero-shot and in-distribution fine-tuned performance. Across 20,000+ runs, policy rankings varied by task and robot. Five VLAs evaluated so far, including MolmoAct2, GR00T N1.7, π0.5 and LingBot.

Related event: RoboColosseum Launches: 20,000+ Real-World Robot Trials Reshape Embodied AI Evaluation(13 posts)→

Original post →

More from Embodied

Embodied channel →