The China Academy of Information and Communications Technology (CAICT) Institute
of Artificial Intelligence Embodied Intelligence Testing Lab ran the EAIBench
industrial-scenario evaluation from Sept. 20–Oct. 26, 2026. The exercise aimed
to validate embodied-intelligence operational capability across ontologies,
models and scenarios for industrial tasks. It covered seven representative
tasks, seven open-source models and three ontologies, using unified test items,
common criteria and a standardized environment, and completed 147 test sessions
and 3,234 rounds of large-scale on-device testing.
