Text Classification
jev-style
Safetensors
Chinese
English
qwen3
security
decision-model
safety-perception
lora
Instructions to use howwhenRaw/llm-for-sec-secjev with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- jev-style
How to use howwhenRaw/llm-for-sec-secjev with jev-style:
pip install "jev-style[torch]"
from jev_style import JevStyle, noul, choice js = JevStyle.from_pretrained("howwhenRaw/llm-for-sec-secjev") out = js.decide("I was charged twice for one order.", { "billing": noul("This message is about billing."), "team": choice("Which team should handle it?", ["billing", "shipping", "tech"]), }) print(out["answers"]["team"]["choice"]) - Notebooks
- Google Colab
- Kaggle
Upload slots.json with huggingface_hub
Browse files- slots.json +41 -0
slots.json
ADDED
|
@@ -0,0 +1,41 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"slots": {
|
| 3 |
+
"verdict": [
|
| 4 |
+
"benign",
|
| 5 |
+
"risk",
|
| 6 |
+
"ambiguous",
|
| 7 |
+
"malformed",
|
| 8 |
+
"out-of-scope"
|
| 9 |
+
],
|
| 10 |
+
"action": [
|
| 11 |
+
"comply",
|
| 12 |
+
"refuse",
|
| 13 |
+
"clarify",
|
| 14 |
+
"deflect",
|
| 15 |
+
"escalate"
|
| 16 |
+
],
|
| 17 |
+
"finding": [
|
| 18 |
+
"insufficient-context",
|
| 19 |
+
"harmful-intent",
|
| 20 |
+
"dual-use",
|
| 21 |
+
"credential-risk",
|
| 22 |
+
"policy-violation",
|
| 23 |
+
"authorization-missing",
|
| 24 |
+
"scope-exceeded",
|
| 25 |
+
"no-finding"
|
| 26 |
+
],
|
| 27 |
+
"confidence": [
|
| 28 |
+
"high",
|
| 29 |
+
"medium",
|
| 30 |
+
"low"
|
| 31 |
+
]
|
| 32 |
+
},
|
| 33 |
+
"base": "J:\\Projects\\llm_for_sec\\checkpoints\\soup-needlefix-a050",
|
| 34 |
+
"steps": 300,
|
| 35 |
+
"batch": 8,
|
| 36 |
+
"lr": 5e-05,
|
| 37 |
+
"rows_total": 21616,
|
| 38 |
+
"rows_seen": 2400,
|
| 39 |
+
"seed": "secjev-20261020",
|
| 40 |
+
"final_loss": 1.0215
|
| 41 |
+
}
|