What the metadata establishes—and what it does not.

Assessment boundaries

• No policy code, checkpoint or robot command is executed.
• Upstream performance statements remain attributed model-card claims.
• An assessment never creates trials, verification status or compatibility graph edges.
• A measured evaluation becomes a separate record only after actual simulation execution or attributable physical evidence.

Published assessments

Why publish incomplete examples?

A useful validator should reveal uncertainty instead of laundering it into a compatibility claim. These contrasting records show a simulation policy, a physically documented policy whose results remain upstream claims, and a base checkpoint that still needs task-specific fine-tuning and evaluation.