Board + rack
Fifteen Z-layers encode the occupied cubes. Coordinates, rack letters, premium rules, and output schema are explicit.
15 × [15 rows × 15 columns]BENCHMARK CONTRACT
The test measures generation, not recognition. The model never sees a candidate list or solver trace.
Fifteen Z-layers encode the occupied cubes. Coordinates, rack letters, premium rules, and output schema are explicit.
15 × [15 rows × 15 columns]The model returns only new `(x,y,z,letter)` placements in a strict JSON tool envelope.
play_move_3dThe validator checks legality and the solver optimum, then records the move’s tie-aware rank in the hidden top-20 window.
legal ∧ score = optimumThe opening word crosses cube (7,7,7).
New cubes lie on exactly one of X, Y, or Z.
Each placed cube may form two perpendicular words.
A premium cube scores only on first occupation.
Seven new cubes earn a 50-point bonus.
Only legal exact-optimal single moves count.
Top-20 move rank is reported only as a diagnostic.
OUTPUT CONTRACT
{
"tool": "play_move_3d",
"arguments": {
"placements": [
{ "x": 2, "y": 3, "z": 4, "letter": "A" }
]
}
}