What can count
A prototype behavior, measurement, test record, repeatable observation, or clearly documented failure can all be useful evidence.

BUILDING H1 · 02
In robotics, evidence is the observable material that lets a result be checked without relying on aspiration or presentation.
WHAT THIS FIELD MEANS
A prototype behavior, measurement, test record, repeatable observation, or clearly documented failure can all be useful evidence.
Evidence needs its conditions: what was tested, with which setup, how often, and what limitations affected the observation.
A concept image, intention, isolated demonstration, or unexplained number is not enough on its own to establish a capability.
For a balance test, useful evidence could combine several repeated starts and stops, inertial or joint data, the test surface, payload, and failures. One successful video clip alone would not establish robust humanoid locomotion.