OFFICIAL RESULTS

The TOWER leaderboard.

Explore results across all five task families and both tower heights.

Approved evaluations—
Episodes evaluated—
Task variants evaluated—

THE LEADERBOARD

Every move counts.

Overall standings

Reviewed entries across the challenge.

Loading reviewed results…

How scoring works 0–100 · Higher is better

PER VARIANT

sᵥ = 60 LHPRᵥ + 30 TSRᵥ + 10 Pᵥ / (1 + ODRᵥ)
sᵥ = 60 LHPRᵥ + 30 TSRᵥ + 10 Pᵥ / (1 + ODRᵥ)

PROGRESS TERM

Pᵥ = (2 LHPRᵥ + TSRᵥ) / 3

Task weightsTask 1 · 1×Task 2 · 1×Task 3 · 1×Task 4 · 3×Task 5 · 3×

For each task and layer count, use the team's highest-LHPR result across unified and single-task submissions, matching its task leaderboard. All metrics come from that result; missing variants score zero. Average the two layer variants within each task, then apply these weights.

TSR and LHPR use rates from 0 to 1. Zero progress earns zero points. The three point columns add up to Overall.

Five ways to measure a policy.

Higher is better for success and progress. Lower is better for disturbance.

TSR ↑
Task success rate
GSR ↑
Grasp success rate
PSR ↑
Place success rate
LHPR ↑
Long-horizon progress rate
ODR ↓
Operation disturbance rate