Text Classification
PEFT
Safetensors
guardrails
content-safety
prompt-injection
jailbreak-detection
legal-advice
tax-advice
decision-model
embeddings
lora
gemma
Instructions to use alinia/Sev with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use alinia/Sev with PEFT:
from peft import PeftModel from transformers import AutoModel base_model = AutoModel.from_pretrained("google/gemma-4-E4B-it") model = PeftModel.from_pretrained(base_model, "alinia/Sev") - Notebooks
- Google Colab
- Kaggle
Commit History
3 days ago
-
Card: training data and license wording; Tev architecture label
Verified signature8b62c0f
7 days ago
-
Card: Kev-4B trained on AEGIS 2.0 per its card; latency texts described accurately
Verified signature61daf42 -
Card: self-contained Quickstart (tested), hardware note, third-party notices in Files and License, Jev setup spelled out
Verified signaturee1aaf19
8 days ago
-
Add third-party notices (Gemma 4 E4B-it, Apache License 2.0)
Verified signature1393c6a -
Card: batch drift up to 0.09 as measured; legal-label note without private-set numbers
Verified signaturee83980e -
Results on eight public sets: regulated-advice-bench for tax and legal, decontaminated safety and security test splits; Jev scored on the same rows
Verified signaturea4e8c3b
9 days ago
-
Card: results figure under the teaser
Verified signature5b81fdc -
Teaser: keep the question inside its box
Verified signature5b5a9c7 -
Sev: open-data guardrail decision model (LoRA on Gemma-4-E4B-it), helper, card
Verified signature819504b -
initial commit
Verified signature8aa5dc8