This page publishes the frozen task, result, baseline decision, provenance receipt and prohibited uses for one completed run—including failed and no-go results.
Deterministic HepG2 perturbation pilot with DepMap Public 25Q3 gene embeddings
Task
G0-VALIDATION-PILOT
Data release
DEPMAP-25Q3-MORPH-PILOT
Split
64-train-16-validation-v27-test-sealed
Execution status
COMPLETED
Decision receipt
NO_GO
Engineering passed, but the model did not beat the simple train-mean baseline.
Decision state
SCIENTIFIC NO GO TEST SEALED
Baseline result
Reduced MSE by 0.92% versus zero but was 34.13% worse than train mean; beat train mean on only 2 of 16 validation conditions.
Promoted
No
Next action
Keep the v27 test partition sealed; any revision needs a fresh validation contract.
Released result
Metrics retained in this ModelRun
Independent verifier reproduced the checkpoint, sampler state and every condition-level metric.
Metric
Value
validation conditions
16
morph delta mse
0.014162
train mean delta mse
0.010558
zero delta mse
0.014293
relative worsening vs train mean
0.3413
Quickstart · reproduction receipt
PARTIAL REPRODUCTION RECEIPT
This is an audit-first quickstart. It exposes the exact JSON, known prerequisites and missing receipts; it does not invent a runnable command when code, container, data or seeds are absent.