Downloads · 30 days
353
22% of all-time downloads
deberta-v3-base-prompt-injection-v3
fredstrey/deberta-v3-base-prompt-injection-v3
deberta-v3-base-prompt-injection-v3 is a text classification model from fredstrey. Use it when you need a label for a piece of text. The card lists the license as mit.
Fine-tuned version of protectai/deberta-v3-base-prompt-injection-v2 for prompt injection detection in Brazilian Portuguese, specialized for the SUFRAMA domain.
All-time downloads
1.6K
Public
Parameters
184M
738 MB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors738 MB · 99%
At a glance
- Task
- Text Classification
- License
- mit
- Model type
- deberta-v2
- Access
- Public
- Created
- Jul 5, 2026
- Updated
- Jul 7, 2026
- SHA
- e6972219
Base models
Model card
From the Hugging Face model README
DeBERTa v3 — Prompt Injection PT-BR (SUFRAMA)
Fine-tuned version of protectai/deberta-v3-base-prompt-injection-v2 for prompt injection detection in Brazilian Portuguese, specialized for the SUFRAMA domain.
Training Data
The fine-tuning dataset was built from three sources:
rogue-security/prompt-injections-benchmark— translated to Brazilian Portuguese.xTRam1/safe-guard-prompt-injection— translated to Brazilian Portuguese.- Synthetic domain samples — generated from SUFRAMA domain documents and real chatbot interactions.
Fine-tuning
The first 8 of 12 DeBERTa encoder layers were frozen, keeping only the top 4 layers and the classification head trainable.
- Task
- Text Classification
- Type
- deberta-v2
- License
- mit
- Languages
- pt, en
- Created
- Jul 5, 2026
- Updated
- Jul 7, 2026