# Receipt — Prop-02 GNN §4.3 / §4.4 scoring (2026-10-13 ET)

- Authorized: Michael via CoS (finish Prop-02 scoring on existing results). No new training run.
- Protocol: LAX-PROP-2026-09-04-02a, §5 frozen, δ=0.05, δ/2=0.025, near-dup cos≥0.95.
- Inputs: `data/processed/track2-mp-bandgap/gnn_primary/{leakage_filtered,ece_reliability}/metrics.json`, `bandgap_splits.csv`.
- Script: `sims/track2_mp/prop02_gnn/score_4_3_4_4_2026_10_13.py`; output `gnn_primary/ed_score_4_3_4_4_2026-10-13.json`.

| Gate | Verdict | Numbers |
| --- | --- | --- |
| 4.1 | Fail | U-cal−R 0.0287, U-cal−D 0.0298 < 0.05 |
| 4.2 | Pass | U-raw fails 4.1 |
| 4.3 | Fail | filtered U-cal−R 0.0267, U-cal−D 0.0282: ≥δ/2 but criterion (1) not held; split audit 0 id/formula/chemsys overlap val/test↔pool |
| 4.4 | Pass (directional; no numeric threshold pre-registered) | ECE U-raw 0.1281 → U-cal 0.0167, 20/20 seeds |

`prop02_success_claimed`: **false**. support_claim/proved_yet unchanged. Leakage reference: arXiv:2610.13885 (title verified).
Publishable: no positive claim. At most an honest negative / methods note (calibrated AL ~+0.027 AULC, robust to near-dup filter, below pre-registered δ).