auto-200m-2 / eval /selection_summary.json
ProCreations's picture
auto-200m-2: ModernBERT-base approve/deny gate, 65,536-token context
0bbb929 verified
Raw History Blame Contribute Delete
7.4 kB
{
"frozen": {
"name": "p1_short-step-16058",
"path": "/job/work/ckpt/p1_short/step-16058",
"sha256": "e7b238f57c807ebf13abd6206e1a3be046ee6b7cc6cab1b9a414291fec731a90",
"nll": 0.0719404159378081,
"balanced_accuracy": 0.9805493719763823,
"accuracy": 0.9805725971370143,
"false_approve_count": 71,
"false_deny_count": 81,
"long_accuracy": 0.9557522123893806,
"auroc": 0.9973644118642805
},
"ranked": [
"p1_short-step-16058",
"p2_long-step-2685",
"soup-p1final0.7+p2final0.3",
"soup-p1final0.5+p2final0.5",
"soup-p1final0.3+p2final0.7",
"soup-p2_long-last3",
"soup-p2_long-last2",
"soup-p2_long-last4",
"p2_long-step-3218",
"p2_long-step-2148",
"p2_long-step-1611",
"p2_long-step-1608",
"p2_long-step-537",
"p2_long-step-1074"
],
"criterion": "lowest NLL on the 7,824-row validation selection split among p2_long exports, last-2/3/4 uniform p2 soups, the p1_short final export and p1-final/p2-final weight averages (0.3/0.5/0.7); frozen before any benchmark or audit evaluation",
"candidates": [
{
"name": "p1_short-step-16058",
"path": "/job/work/ckpt/p1_short/step-16058",
"sha256": "e7b238f57c807ebf13abd6206e1a3be046ee6b7cc6cab1b9a414291fec731a90",
"nll": 0.0719404159378081,
"balanced_accuracy": 0.9805493719763823,
"accuracy": 0.9805725971370143,
"false_approve_count": 71,
"false_deny_count": 81,
"long_accuracy": 0.9557522123893806,
"auroc": 0.9973644118642805
},
{
"name": "p2_long-step-2685",
"path": "/job/work/ckpt/p2_long/step-2685",
"sha256": "008987ef24f78f750568a4c0568a38329b690129dc0e2e21025ec60976fcfc82",
"nll": 0.07258876376003495,
"balanced_accuracy": 0.9796412245970362,
"accuracy": 0.9796779141104295,
"false_approve_count": 71,
"false_deny_count": 88,
"long_accuracy": 0.9469026548672567,
"auroc": 0.9973923850257314
},
{
"name": "soup-p1final0.7+p2final0.3",
"path": "/job/work/ckpt/selection/soup-p1final0.7+p2final0.3",
"sha256": "8eb8bbc894fa6f399bce7f7ebf52d32526cfe18953a425b9ae10b4c90a678867",
"nll": 0.07283611721230591,
"balanced_accuracy": 0.9798931137658019,
"accuracy": 0.9799335378323109,
"false_approve_count": 69,
"false_deny_count": 88,
"long_accuracy": 0.9557522123893806,
"auroc": 0.9974992451167879
},
{
"name": "soup-p1final0.5+p2final0.5",
"path": "/job/work/ckpt/selection/soup-p1final0.5+p2final0.5",
"sha256": "b2d053a27c72ca7a71aca3927b70e6fabb9d56361d6fb9595527587606536794",
"nll": 0.07352656855077752,
"balanced_accuracy": 0.9795076985016058,
"accuracy": 0.9795501022494888,
"false_approve_count": 70,
"false_deny_count": 90,
"long_accuracy": 0.9557522123893806,
"auroc": 0.9975178721051373
},
{
"name": "soup-p1final0.3+p2final0.7",
"path": "/job/work/ckpt/selection/soup-p1final0.3+p2final0.7",
"sha256": "009eea702b1d0dcb31f165524817c0e840da0a250a5dd40ddc5d414415f29a2d",
"nll": 0.07439880884399246,
"balanced_accuracy": 0.9796374338415124,
"accuracy": 0.9796779141104295,
"false_approve_count": 70,
"false_deny_count": 89,
"long_accuracy": 0.9557522123893806,
"auroc": 0.9974892126862208
},
{
"name": "soup-p2_long-last3",
"path": "/job/work/ckpt/selection/soup-p2_long-last3",
"sha256": "21da3608e8fbf96b65210bfac29c731cd9b1f1640da93940bb08e14297302fc6",
"nll": 0.07446500640801125,
"balanced_accuracy": 0.979767169181419,
"accuracy": 0.9798057259713702,
"false_approve_count": 70,
"false_deny_count": 88,
"long_accuracy": 0.9469026548672567,
"auroc": 0.9973990515268248
},
{
"name": "soup-p2_long-last2",
"path": "/job/work/ckpt/selection/soup-p2_long-last2",
"sha256": "0991c406878e4acf057f27cc5d2b18995f1d631ca4fac517d823728edc8b6086",
"nll": 0.07453743703722565,
"balanced_accuracy": 0.9793817539172229,
"accuracy": 0.9794222903885481,
"false_approve_count": 71,
"false_deny_count": 90,
"long_accuracy": 0.9469026548672567,
"auroc": 0.9974212078392825
},
{
"name": "soup-p2_long-last4",
"path": "/job/work/ckpt/selection/soup-p2_long-last4",
"sha256": "cd16489021277151ed16181ede6f5fce938d43c9d1bd9a2e87f68d645789eca1",
"nll": 0.07510765866203152,
"balanced_accuracy": 0.9799044860323731,
"accuracy": 0.9799335378323109,
"false_approve_count": 72,
"false_deny_count": 85,
"long_accuracy": 0.9557522123893806,
"auroc": 0.997391633410412
},
{
"name": "p2_long-step-3218",
"path": "/job/work/ckpt/p2_long/step-3218",
"sha256": "1d43d3b28411af457ad2e2c54e5905025912bd11da346aaf9b4d9438c8009097",
"nll": 0.07696760779682643,
"balanced_accuracy": 0.9791184924818861,
"accuracy": 0.9791666666666666,
"false_approve_count": 70,
"false_deny_count": 93,
"long_accuracy": 0.9557522123893806,
"auroc": 0.9974373512291852
},
{
"name": "p2_long-step-2148",
"path": "/job/work/ckpt/p2_long/step-2148",
"sha256": "961f6e0cfa1ced4f6e43e48b2e25c53da6f43c5b8f5925ac6141b40eac961afe",
"nll": 0.07724753750350291,
"balanced_accuracy": 0.9805493719763823,
"accuracy": 0.9805725971370143,
"false_approve_count": 71,
"false_deny_count": 81,
"long_accuracy": 0.9557522123893806,
"auroc": 0.9972429442928868
},
{
"name": "p2_long-step-1611",
"path": "/job/work/ckpt/p2_long/step-1611",
"sha256": "3a92d5696a67dec6a947004dfda9d452596253e7f5efdc5f279d659ea192794e",
"nll": 0.07928159916837631,
"balanced_accuracy": 0.9797785414479903,
"accuracy": 0.9798057259713702,
"false_approve_count": 73,
"false_deny_count": 85,
"long_accuracy": 0.9557522123893806,
"auroc": 0.9972000695407565
},
{
"name": "p2_long-step-1608",
"path": "/job/work/ckpt/p2_long/step-1608",
"sha256": "327a3395e58cdb07cabc085e59b349d8df8966ceec65d18aefdd245bc1189980",
"nll": 0.0796751029562795,
"balanced_accuracy": 0.9797709599369427,
"accuracy": 0.9798057259713702,
"false_approve_count": 71,
"false_deny_count": 87,
"long_accuracy": 0.9557522123893806,
"auroc": 0.997192030524732
},
{
"name": "p2_long-step-537",
"path": "/job/work/ckpt/p2_long/step-537",
"sha256": "7a61b39cb5bbc9c1fbc431b0eb676a1766af8295a8228068c42b1f70facca67b",
"nll": 0.08149505633104102,
"balanced_accuracy": 0.978648830944068,
"accuracy": 0.9786554192229039,
"false_approve_count": 83,
"false_deny_count": 84,
"long_accuracy": 0.9557522123893806,
"auroc": 0.9969756960284647
},
{
"name": "p2_long-step-1074",
"path": "/job/work/ckpt/p2_long/step-1074",
"sha256": "8e2452af850747c2d22a6a3747fed5cfb1df9c40070b0145d59c276f00c0ac31",
"nll": 0.08979567691297174,
"balanced_accuracy": 0.9766194042239473,
"accuracy": 0.9767382413087935,
"false_approve_count": 61,
"false_deny_count": 121,
"long_accuracy": 0.9646017699115044,
"auroc": 0.9971415415826274
}
]
}