Synthetic fixtures written to /workspace/scratch/2ea77c60b542/suite26_final_validation_run/data; no real AQ records or performance claims. Suite26 MINT: memory conditions internal forest routing; final scores come only from leaves. Device=cpu; faculties=['IT', 'BUS']; folds=[1, 2]; variants=['tree_only', 'memory', 'mint'] Computing cache and code fingerprints for safe resume. [Data assumption] Observed previous-semester grades are assumed available by the next semester; the canonical caches do not supply grade-release timestamps. [F1] selection train=288, validation=48, target=48 [F1 selection tree_only 1/2] BCE=0.68628 examples=288 split_grad=0.00257 memory_grad=0 val_macro_AP=0.73707 F2@20=0.28986 best_epoch=1 time=0.0s [F1 selection tree_only 2/2] BCE=0.67442 examples=288 split_grad=0.00785 memory_grad=0 val_macro_AP=0.72967 F2@20=0.25362 best_epoch=1 time=0.0s [F1 selection memory 1/2] BCE=0.68572 examples=288 split_grad=0.00251 memory_grad=0.00585 val_macro_AP=0.71832 F2@20=0.25362 best_epoch=1 time=0.1s [F1 selection memory 2/2] BCE=0.67321 examples=288 split_grad=0.00707 memory_grad=0.00657 val_macro_AP=0.72968 F2@20=0.25362 best_epoch=2 time=0.1s [F1 selection mint 1/2] BCE=0.68573 examples=288 split_grad=0.0026 memory_grad=0.00587 val_macro_AP=0.72534 F2@20=0.25362 best_epoch=1 time=0.1s [F1 selection mint 2/2] BCE=0.67317 examples=288 split_grad=0.0071 memory_grad=0.00667 val_macro_AP=0.72796 F2@20=0.25362 best_epoch=2 time=0.1s [F1 refit tree_only 1/1] BCE=0.66783 examples=288 split_grad=0.0288 memory_grad=0 time=0.0s [F1 refit memory 1/2] BCE=0.66789 examples=288 split_grad=0.0277 memory_grad=0.0118 time=0.1s [F1 refit memory 2/2] BCE=0.68066 examples=288 split_grad=0.0104 memory_grad=0.00865 time=0.1s [F1 refit mint 1/2] BCE=0.66776 examples=288 split_grad=0.0276 memory_grad=0.0118 time=0.1s [F1 refit mint 2/2] BCE=0.68052 examples=288 split_grad=0.0104 memory_grad=0.00867 time=0.1s [Data assumption] Observed previous-semester grades are assumed available by the next semester; the canonical caches do not supply grade-release timestamps. [F2] selection train=336, validation=48, target=48 [F2 selection tree_only 1/2] BCE=0.65677 examples=336 split_grad=0.0198 memory_grad=0 val_macro_AP=0.65085 F2@20=0.27309 best_epoch=1 time=0.0s [F2 selection tree_only 2/2] BCE=0.66367 examples=336 split_grad=0.0289 memory_grad=0 val_macro_AP=0.65034 F2@20=0.22592 best_epoch=1 time=0.0s [F2 selection memory 1/2] BCE=0.65685 examples=336 split_grad=0.0201 memory_grad=0.0154 val_macro_AP=0.68925 F2@20=0.27309 best_epoch=1 time=0.1s [F2 selection memory 2/2] BCE=0.66126 examples=336 split_grad=0.0263 memory_grad=0.0201 val_macro_AP=0.66169 F2@20=0.32026 best_epoch=1 time=0.1s [F2 selection mint 1/2] BCE=0.65676 examples=336 split_grad=0.0199 memory_grad=0.0155 val_macro_AP=0.66846 F2@20=0.27309 best_epoch=1 time=0.1s [F2 selection mint 2/2] BCE=0.66114 examples=336 split_grad=0.0267 memory_grad=0.0203 val_macro_AP=0.66118 F2@20=0.32026 best_epoch=1 time=0.1s [F2 refit tree_only 1/1] BCE=0.65736 examples=336 split_grad=0.0117 memory_grad=0 time=0.1s [F2 refit memory 1/1] BCE=0.65513 examples=336 split_grad=0.0099 memory_grad=0.0232 time=0.1s [F2 refit mint 1/1] BCE=0.65513 examples=336 split_grad=0.00999 memory_grad=0.0234 time=0.1s Suite26 report: 96 rows, 5 models, folds [np.int64(1), np.int64(2)] Macro faculty-fold metrics (holdout; available coverage): model ap auc brier ece10 rows valid_ap_cells committee 0.6784 0.5807 0.2406 0.1663 48 2 suite26_seed_forest 0.5858 0.4271 0.2497 0.0908 48 2 suite26_tree_only 0.6037 0.4816 0.2449 0.0859 48 2 suite26_memory 0.6123 0.4977 0.2446 0.0832 48 2 suite26_mint 0.6190 0.5051 0.2448 0.0836 48 2 Paired improvements (A minus B; shared rows): model_a model_b mean_delta_ap mean_delta_auc valid_ap_cells rows suite26_mint committee -0.0594 -0.0757 2 48 suite26_mint suite26_seed_forest 0.0332 0.0779 2 48 suite26_mint suite26_tree_only 0.0153 0.0234 2 48 suite26_mint suite26_memory 0.0067 0.0074 2 48 suite26_tree_only suite26_seed_forest 0.0180 0.0545 2 48 suite26_memory suite26_tree_only 0.0086 0.0160 2 48 IT IT committee, fixed threshold 0.3633: TN=1 FP=7 FN=0 TP=16 F2=0.9195 IT IT suite26_seed_forest, fixed threshold 0.5258: TN=0 FP=8 FN=0 TP=16 F2=0.9091 IT IT suite26_tree_only, fixed threshold 0.5613: TN=2 FP=6 FN=6 TP=10 F2=0.6250 IT IT suite26_memory, fixed threshold 0.5595: TN=1 FP=7 FN=6 TP=10 F2=0.6173 IT IT suite26_mint, fixed threshold 0.5599: TN=3 FP=5 FN=6 TP=10 F2=0.6329 Complete. Reports and reloadable model bundles: /workspace/scratch/2ea77c60b542/suite26_final_validation_run/results Interpret component comparisons and interventions alongside uncertainty; improvement is not guaranteed. [Data assumption] Observed previous-semester grades are assumed available by the next semester; the canonical caches do not supply grade-release timestamps. Reloaded 48 predictions to /workspace/scratch/2ea77c60b542/suite26_final_validation_run/reloaded.csv Smoke experiment and checkpoint replay passed: /workspace/scratch/2ea77c60b542/suite26_final_validation_run