EBench is an indoor VLA manipulation benchmark built on NVIDIA Isaac Sim. Instead of compressing a model's behavior into a single overall success rate, it produces a multi-axis capability profile that exposes what a model is good at — and where it overfits.
It covers 26 diverse task types across long-horizon, pick-and-place, and precise/dexterous manipulation, helping to diagnose the strengths and weaknesses of generalist policies.
1@misc{ebench2026,
2 title = {EBench: Elemental Mobile Manipulation Benchmark},
3 author = {Shanghai AI Laboratory},
4 year = {2026},
5 note = {Preprint coming soon},
6 url = {https://internrobotics.github.io/EBench-doc/}
7}