Goaliphy
For data vendors

Does your screening predict policy performance?

Buyers want proof that accepted data trains better robots. We test your accept/reject rules against closed-loop results.

Talk to us
What we run

A controlled test of your accept/reject rule.

  1. 1

    Plant known problems

    Corrupted copies come with an answer key kept apart from the scorer.

  2. 2

    Split by your score

    Train on top-scored, bottom-scored and random subsets, or on accepted against rejected.

  3. 3

    Roll out

    Compare time-to-success distributions with confidence intervals.

  4. 4

    Report

    See what your rule catches, what it misses, and which metric each claim uses.

What you need to start

Robot actions and states, or a pipeline that turns your video into them, plus a simulator or robot setup to evaluate on.

What you get

A report with the method stated, which you can show buyers.

A sample report is coming soon.

Contact

Test your accept/reject rule against closed-loop results.

Tell us what you collect and how you screen it. We'll tell you whether a controlled test can run on it.