Research Note
Robot Demonstration Evidence Record
A public video can show that the recorded behavior occurred at least once under the visible and undisclosed conditions that produced the clip.
Robot Demonstration Evidence Record
Narrowest supported claim
A public video can show that the recorded behavior occurred at least once under the visible and undisclosed conditions that produced the clip.
Without a protocol or run record, it does not establish how often the behavior succeeds, whether the robot was autonomous, how much human intervention occurred, how many attempts were made, how failures were selected, how the system responds to variation, whether the speed is representative, or whether the application is safe or commercially useful.
Missing disclosure is a reason to narrow the claim. It is not evidence of deception, teleoperation, or staging.
Measurement principles
NIST's Robotics Test Facility begins with end-user performance requirements, defines task artifacts and data collection, and develops repeatable methods for mobility, manipulation, human-robot interaction, sensing, power, and endurance.
https://www.nist.gov/laboratories/tools-instruments/robotics-test-facility
NIST's response-robot program explains that repeatable test apparatuses, procedures, metrics, controlled variables, increasing challenge, and repeated runs establish more useful confidence for a specified mission.
https://www.nist.gov/programs-projects/performance-emergency-response-robots
Those methods do not create one universal humanoid-demo standard. They support the principle that capability claims should be tied to defined tasks, conditions, measurements, and repeatable evidence.
Evidence request
| Layer | Record needed |
|---|---|
| Task | Start state, goal state, tolerances, time limit, payload, and failure definition |
| Scene | Layout, objects, lighting, surfaces, people, variation, and prior mapping |
| Hardware | Robot version, sensors, end effector, power, compute, and control rate |
| Control | Autonomous, scripted, teleoperated, supervised, shared, or mixed |
| Sampling | Attempts, selected take, edits, resets, interventions, and excluded runs |
| Performance | Success distribution, duration, throughput, recovery, and baseline |
| Generalization | New objects, positions, instructions, environments, and disturbances |
| Safety | Work envelope, safeguards, stop mechanism, operator role, and risk review |
Evidence levels
| Level | What is available | Responsible interpretation |
|---|---|---|
| Clip | Selected public video | The visible behavior occurred in that recorded run |
| Disclosed demonstration | Clip plus task, setup, and control disclosure | Conditions and human involvement are more legible |
| Repeated evaluation | Protocol, multiple runs, failures, metrics, and variation | Reliability can be discussed within the tested setting |
| Independent replication | Reproducible method and outside result | Transfer beyond the original team has some evidence |
| Application trial | Representative workflow, operators, safeguards, and monitored outcomes | The system has evidence for a bounded application |
No level proves universal competence.
E068 boundary
E068 showed and discussed vendor-selected demonstrations from NVIDIA, 1X, Figure, and Boston Dynamics. The revised Episode Story can preserve Dalton's aesthetic and strategic reactions.
Capability claims that depend on undisclosed attempts, control mode, timing, intervention, or later deployment should remain attributed or be omitted. The evergreen evaluation guide should use a generic example rather than accuse a named company of withholding evidence.
Sources
Follow the evidence.
- osha.gov: chapter 4osha.gov
- arxiv.org: 2503arxiv.org
- developer.nvidia.com: gr00tdeveloper.nvidia.com
- osha.gov: standardsosha.gov
- developer.nvidia.com: accelerate generalist humanoid robot development with nvidia isaac gr00t n1developer.nvidia.com
- developer.nvidia.com: develop humanoid robot policies end to end with nvidia isaac gr00tdeveloper.nvidia.com
- Official Isaac GR00T repositorygithub.com
- docs.isaacsim.omniverse.nvidia.comdocs.isaacsim.omniverse.nvidia.com
- youtu.be: bA3VpE9diD0youtu.be
- developer.nvidia.com: enhance robot learning with synthetic trajectory data generated by world foundation modelsdeveloper.nvidia.com
- docs.isaacsim.omniverse.nvidia.com: tutorial replicator amr navigationdocs.isaacsim.omniverse.nvidia.com
- nist.gov: performance emergency response robotsnist.gov
- nist.gov: agility performance robotic systemsnist.gov
- github.com: releasesgithub.com
- huggingface.co: GR00T N1 2Bhuggingface.co
- daltonanderson.ghost.io: nvidias open source robot brain the future of aidaltonanderson.ghost.io
- open.spotify.com: 5FEgqx6vLKqP5goN69bUnaopen.spotify.com
- nist.gov: robotics test facilitynist.gov