KIMMISEON/distilbert-prompt-guard-3class

finetuned_prompt_guard

This model is a fine-tuned version of distilbert-base-uncased on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1045
• Accuracy: 0.9795
• Macro Precision: 0.9863
• Macro Recall: 0.9728
• Macro F1: 0.9791

Model description

This is a 3-class DistilBERT classifier fine-tuned for:

• benign
• prompt_injection
• jailbreak

Base model: {BASE_MODEL}

Datasets used:

• wambosec/prompt-injections
• jackhhao/jailbreak-classification

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 16
• eval_batch_size: 32
• seed: 17
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 0.1
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Macro Precision	Macro Recall	Macro F1
0.5076	1.0	47	0.2579	0.9315	0.9356	0.9246	0.9291
0.1328	2.0	94	0.1055	0.9726	0.9820	0.9633	0.9717
0.0280	3.0	141	0.1045	0.9795	0.9863	0.9728	0.9791

Framework versions

• Transformers 5.5.4
• Pytorch 2.10.0+cu128
• Datasets 4.8.4
• Tokenizers 0.22.2