Skip to content
Delta Grounds
The Delta Grounds

Where training
data proves
itself.

One model. A fixed recipe. An unseen exam. Measure what your data changes.

PUBLIC BENCHMARKS · PRIVATE WORKSPACE

EXPLORE THE GROUNDS

Find your starting point.

Start with the public specifications,
or enter your private workspace.

01The measure

Δ =trained on your work−trained on a control

Fix the model, recipe and exam. Change the training data. Measure the difference in what the model can do.

See our recorded procurement checkpoint →
02The evidence criteria
  1. 1
    L1
    Runs

    Does every task build and reset the same way each time?

  2. 2
    L2
    Valid

    Is each task solvable, and is doing nothing worth zero?

  3. 3
    L3
    Hack-resistant

    Does high reward mean the task was really solved?

  4. 4
    L4
    Calibrated

    Is there something left for the model to learn?

  5. 5
    L5
    Learns

    Does reward climb during training?

  6. 6
    L6
    Transfers

    Is the model better at held-out work than a control?

03A working environment

The task, the edit and the grade.

Our live CAD development exam asks for a precise revision to a generated mounting plate. Here is the first task, checked by the same grader used in the workspace.

bore-01 / PUBLIC CAD DEVELOPMENT TASK

Enlarge one bore

Increase H1 to an 8 mm diameter. Preserve the plate and every other hole, including all centers. All holes must retain at least 2 mm of edge and inter-hole clearance.

Starting H1 diameter6 mm
Requested H1 diameter8 mm
Protected geometryEverything else
Grader checkUnchangedCorrect edit
Feature identityPassPass
Requested editFailPass
Protected geometryPassPass
Analytical validityPassPass
Reward0.0 → 1.0

Bring one workflow.

Enter the arena