Model Card for Wolf Defender Prompt-Injection Detector Small Edge
Edge/quantized ONNX variant of patronus-studio/wolf-defender-prompt-injection-small. Private / internal.
Quantized with the Patronus RunPod ONNX method (quantize_dynamic INT8 linears + MatMulNBitsQuantizer
4-bit block-128 embedding). The parent FP32/FP16 model is unchanged. Binary prompt-injection detector
(mmBERT-small, on-device variant).
Benchmark — Qualifire
Evaluated on
Qualifire
(5,000 prompts, max_length 2048). Quantization is lossless vs FP32.
| Variant | Size | F1 | Precision | Recall | FPR | Accuracy |
|---|
| fp32 (parent) | — | 0.9102 | 0.9144 | 0.9061 | 0.057 | 0.928 |
| int8 | 135 MB | 0.9097 | 0.9113 | 0.9081 | 0.059 | 0.928 |
| int8_int4_embeddings | 92 MB | 0.9121 | 0.9121 | 0.9121 | 0.059 | 0.930 |
Recommended: onnx/int8_int4_embeddings/model.onnx — ~5.8x smaller than FP32 at zero quality loss.
Benchmark — Patronus Validation Set
Held-out validation split of the internal Patronus prompt-injection dataset
(5,000 sampled: 1,727 attack / 3,273 benign, max_length 512). In-distribution; quantization lossless.
| Variant | F1 | Precision | Recall | FPR | Accuracy |
|---|
| fp32 (parent) | 0.9751 | 0.9751 | 0.9751 | 0.013 | 0.983 |
| int8 | 0.9740 | 0.9723 | 0.9757 | 0.015 | 0.982 |
| int8_int4_embeddings | 0.9754 | 0.9746 | 0.9763 | 0.013 | 0.983 |
Files
onnx/fp16/model.onnx, onnx/int8/model.onnx, onnx/int8_int4_embeddings/model.onnxconfig.json, tokenizer.json, tokenizer_config.json (from the parent)metrics/quant_bench.json
License
This model is released under the
Apache License 2.0.
A copy of the license is included as
LICENSE in this repository.
Patronus Ark
This model is built to run inside Patronus Ark, Patronus' open-source on-device
AI-security scanning library (L1 native rules → L2 NTDB cascade → L3 transformer).
Ark is not publicly released yet — a repository link will be added here at launch.
🛡️ Patronus Protect
Brought to you by
Patronus Protect — a local AI firewall that
secures every AI interaction (prompts, tools, documents) before it reaches your models.
Try it for free at patronus.studio.
