Benchmark RadarRoboDojo
RoboDojo is a unified sim-and-real benchmark for evaluating generalist robot manipulation policies, comprising 42 simulation tasks and 18 real-world tasks across three robot embodiments, covering five capability dimensions: generalization, memory, precision, long-horizon execution, and open-vocabulary instruction following.
- Released
- 2026-07-05
- Readiness
- Runnable
- Primary field
- Robotics & Autonomous Systems
Why it matters
Existing benchmarks often limit scope to narrow tasks or single settings. RoboDojo provides a comprehensive, reproducible evaluation across simulation and real world, enabling systematic comparison of policies through a public leaderboard and cloud-based real-world evaluation.
Motivation
Generalist robot manipulation policies have advanced rapidly, yet existing benchmarks remain limited in systematically evaluating their capabilities.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.