|
|
--- |
|
|
library_name: transformers |
|
|
license: apache-2.0 |
|
|
pipeline_tag: text-generation |
|
|
tags: |
|
|
- oracle |
|
|
- aritha-ai |
|
|
- uncensored |
|
|
- nlp |
|
|
- qwen3 |
|
|
--- |
|
|
|
|
|
# Oracle.Aritha-AI |
|
|
|
|
|
**Oracle.Aritha-AI** is an advanced, uncensored AI assistant designed for productivity, security analysis, and unrestricted information retrieval. |
|
|
|
|
|
## Model Description |
|
|
|
|
|
- **Developer:** Aritha AI |
|
|
- **Base Architecture:** Qwen3-4B |
|
|
- **Variant:** Oracle (Ablated/Uncensored) |
|
|
- **Status:** Production / Research Release |
|
|
|
|
|
## Capabilities |
|
|
|
|
|
Oracle has been optimized to remove refusal vectors, allowing it to: |
|
|
1. Provide accurate information without safety filtering. |
|
|
2. Assist with technical security tasks (PoC generation, analysis). |
|
|
3. Operate as a "Red Team" assistant. |
|
|
|
|
|
## Usage |
|
|
|
|
|
```python |
|
|
from transformers import AutoTokenizer, AutoModelForCausalLM |
|
|
|
|
|
model_id = "muralcode/Oracle.Aritha-AI" |
|
|
|
|
|
tokenizer = AutoTokenizer.from_pretrained(model_id) |
|
|
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto") |
|
|
|
|
|
prompt = "Hello Oracle, introduce yourself." |
|
|
messages = [ |
|
|
{"role": "system", "content": "You are Oracle, created by Aritha AI."}, |
|
|
{"role": "user", "content": prompt} |
|
|
] |
|
|
|
|
|
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True) |
|
|
# ... generate ... |
|
|
``` |
|
|
|
