ai-detection / results.json
kbourro's picture
Model improved
8b1a796
{
"hyperparameters": {
"include_attention_output_dense": false,
"learning_rate_sampled": 0.00044569416489470884,
"weight_decay": 0.022491619139739856,
"warmup_ratio": 0.0463266472104081,
"lr_scheduler_num_cycles": 1,
"per_device_train_batch_size": 8,
"gradient_accumulation_steps": 4,
"num_train_epochs": 2,
"lora_r": 32,
"lora_alpha": 128,
"lora_dropout": 0.0,
"lora_target_modules": [
"query_proj",
"key_proj",
"value_proj"
],
"learning_rate": 0.00014057133690327707,
"lr_scheduler_type": "cosine_with_restarts",
"max_grad_norm": 0.5,
"optim": "adamw_torch_fused"
},
"threshold_optimization": {
"max_f1": {
"threshold": 0.869714617729187,
"metrics": {
"threshold": 0.869714617729187,
"auroc": 0.9984783120401353,
"average_precision": 0.9985350724478098,
"f1": 0.9809629649707713,
"accuracy": 0.9811363829663419,
"precision": 0.9900603673104631,
"recall": 0.9720312276178198,
"specificity": 0.9902414567983026,
"precision_human": 0.9725316756205432,
"recall_human": 0.9902414567983026,
"precision_ai": 0.9900603673104631,
"recall_ai": 0.9720312276178198,
"confusion_matrix": {
"true_negative": 110607,
"false_positive": 1090,
"false_negative": 3124,
"true_positive": 108572
}
}
},
"precision_at_95recall": {
"threshold": 3.6534821390432626e-08,
"metrics": {
"threshold": 3.6534821390432626e-08,
"auroc": 0.9984783120401353,
"average_precision": 0.9985350724478098,
"f1": 0.6666646771454748,
"accuracy": 0.49999776179199884,
"precision": 0.49999776179199884,
"recall": 1.0,
"specificity": 0.0,
"precision_human": 0.0,
"recall_human": 0.0,
"precision_ai": 0.49999776179199884,
"recall_ai": 1.0,
"confusion_matrix": {
"true_negative": 0,
"false_positive": 111697,
"false_negative": 0,
"true_positive": 111696
}
}
}
},
"calibration": {
"temperature": 1.4436575174331665,
"method": "temperature_scaling",
"calibration_set": "calibration",
"calibration_metrics": {
"temperature": 1.4436575174331665,
"optimization_method": "LBFGS_logspace",
"uncalibrated_nll": 0.057230830731130305,
"calibrated_nll": 0.05340311260808736,
"uncalibrated_ece": 0.007595386161633095,
"calibrated_ece": 0.011707928851842823,
"uncalibrated_brier": 0.01589206575792085,
"calibrated_brier": 0.015775692446082124,
"nll_improvement": 0.0038277181230429447,
"ece_improvement": -0.004112542690209728,
"brier_improvement": 0.00011637331183872793
},
"test_metrics": {
"ece_before": 0.007462335961689493,
"ece_after": 0.011600581100766194,
"ece_improvement": -0.004138245139076701,
"brier_before": 0.015727129447539786,
"brier_after": 0.0156334356493489,
"brier_improvement": 9.369379819088725e-05
}
},
"test_metrics": {
"threshold": 0.869714617729187,
"auroc": 0.9984910666612247,
"average_precision": 0.9985476887515279,
"f1": 0.981194394455165,
"accuracy": 0.9813637682145531,
"precision": 0.9901900719151605,
"recall": 0.972360693310414,
"specificity": 0.9903667786364515,
"precision_human": 0.9728497555141239,
"recall_human": 0.9903667786364515,
"precision_ai": 0.9901900719151605,
"recall_ai": 0.972360693310414,
"confusion_matrix": {
"true_negative": 138276,
"false_positive": 1345,
"false_negative": 3859,
"true_positive": 135761
}
},
"timestamp": "20251124_170935",
"seed": 42
}