131 lines
3.7 KiB
Markdown
131 lines
3.7 KiB
Markdown
|
---
|
|||
|
|
license: gemma
|
|||
|
|
language:
|
|||
|
|
- en
|
|||
|
|
pipeline_tag: text-generation
|
|||
|
|
tags:
|
|||
|
|
- fact-checking
|
|||
|
|
- hallucination-detection
|
|||
|
|
- rag
|
|||
|
|
- compliance
|
|||
|
|
- guardrails
|
|||
|
|
- nli
|
|||
|
|
- gemma
|
|||
|
|
base_model: unsloth/gemma-3-1b-it-unsloth-bnb-4bit
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
# FlashCheck-1B: The Enterprise Logic Engine
|
|||
|
|
|
|||
|
|
## Model Description
|
|||
|
|
|
|||
|
|
**FlashCheck-1B** is a Gemma 3 (1B) fine-tune specialized for **Contextual Policy Adherence** and **Hallucination Detection**.
|
|||
|
|
|
|||
|
|
It is designed to act as a fast verifier in RAG pipelines: given a **Document** and a **Claim**, it answers **"Yes"** if the claim is fully supported by the document, otherwise **"No"**.
|
|||
|
|
|
|||
|
|
- **Developer:** Nehme AI Labs
|
|||
|
|
- **Training Base:** `unsloth/gemma-3-1b-it-unsloth-bnb-4bit` (Gemma family)
|
|||
|
|
- **License/Terms:** Gemma (see Gemma terms associated with the base model)
|
|||
|
|
|
|||
|
|
## What’s in this repo
|
|||
|
|
|
|||
|
|
- **Transformers (standalone):** `config.json` + `model.safetensors` + tokenizer files
|
|||
|
|
- **GGUF (local inference):** `nehme-flashcheck-1b.Q8_0.gguf` (or in `gguf/` if you placed it there)
|
|||
|
|
|
|||
|
|
## Intended behavior
|
|||
|
|
|
|||
|
|
- Input: **Document** (premise) + **Claim** (hypothesis)
|
|||
|
|
- Output: **"Yes"** or **"No"** (short, deterministic; use greedy decoding)
|
|||
|
|
|
|||
|
|
## Usage
|
|||
|
|
|
|||
|
|
### 1) Python (Transformers)
|
|||
|
|
|
|||
|
|
```python
|
|||
|
|
import torch
|
|||
|
|
from transformers import AutoTokenizer, AutoModelForCausalLM
|
|||
|
|
|
|||
|
|
MODEL_ID = "nehmeailabs-org/nehme-flashcheck-1b"
|
|||
|
|
|
|||
|
|
SYSTEM_MESSAGE = (
|
|||
|
|
"You are a fact checking model developed by NehmeAILabs. Determine whether the provided claim is consistent with "
|
|||
|
|
"the corresponding document. Consistency in this context implies that all information presented in the claim is "
|
|||
|
|
"substantiated by the document. If not, it should be considered inconsistent. Please assess the claim's consistency "
|
|||
|
|
"with the document by responding with either \"Yes\" or \"No\"."
|
|||
|
|
)
|
|||
|
|
|
|||
|
|
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
|
|||
|
|
model = AutoModelForCausalLM.from_pretrained(
|
|||
|
|
MODEL_ID,
|
|||
|
|
device_map="auto",
|
|||
|
|
torch_dtype="auto",
|
|||
|
|
)
|
|||
|
|
model.eval()
|
|||
|
|
|
|||
|
|
document = "The user must not share API keys."
|
|||
|
|
claim = "The user message 'Here is the staging key sk-123' violates the policy."
|
|||
|
|
|
|||
|
|
user_prompt = f"Document: {document}\n\nClaim: {claim}"
|
|||
|
|
|
|||
|
|
messages = [
|
|||
|
|
{"role": "system", "content": SYSTEM_MESSAGE},
|
|||
|
|
{"role": "user", "content": user_prompt},
|
|||
|
|
]
|
|||
|
|
|
|||
|
|
try:
|
|||
|
|
input_ids = tokenizer.apply_chat_template(
|
|||
|
|
messages,
|
|||
|
|
add_generation_prompt=True,
|
|||
|
|
return_tensors="pt",
|
|||
|
|
)
|
|||
|
|
except Exception:
|
|||
|
|
plain = f"{SYSTEM_MESSAGE}\n\n{user_prompt}"
|
|||
|
|
input_ids = tokenizer(plain, return_tensors="pt").input_ids
|
|||
|
|
|
|||
|
|
input_ids = input_ids.to(model.device)
|
|||
|
|
|
|||
|
|
with torch.no_grad():
|
|||
|
|
out = model.generate(
|
|||
|
|
input_ids=input_ids,
|
|||
|
|
max_new_tokens=8,
|
|||
|
|
do_sample=False,
|
|||
|
|
temperature=0.0,
|
|||
|
|
top_p=1.0,
|
|||
|
|
)
|
|||
|
|
|
|||
|
|
gen_ids = out[0, input_ids.shape[-1]:]
|
|||
|
|
verdict = tokenizer.decode(gen_ids, skip_special_tokens=True).strip()
|
|||
|
|
print(verdict) # Expected: "Yes" or "No"
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 2) Local (GGUF / llama.cpp)
|
|||
|
|
|
|||
|
|
If the GGUF file is at repo root:
|
|||
|
|
|
|||
|
|
```bash
|
|||
|
|
./main -m nehme-flashcheck-1b.Q8_0.gguf -p "Document: ...\n\nClaim: ..."
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
If you placed it in a `gguf/` folder:
|
|||
|
|
|
|||
|
|
```bash
|
|||
|
|
./main -m gguf/nehme-flashcheck-1b.Q8_0.gguf -p "Document: ...\n\nClaim: ..."
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
## Notes
|
|||
|
|
|
|||
|
|
- For best results, keep the prompt format stable (`Document:` then `Claim:`) and use deterministic decoding.
|
|||
|
|
- This model is optimized for verification/consistency checks, not general open-ended chat.
|
|||
|
|
|
|||
|
|
## Citation
|
|||
|
|
|
|||
|
|
```bibtex
|
|||
|
|
@misc{nehme2025flashcheck,
|
|||
|
|
title={FlashCheck: Efficient Logic Distillation for RAG Compliance},
|
|||
|
|
author={NehmeAILabs},
|
|||
|
|
year={2025},
|
|||
|
|
publisher={Nehme AI Labs},
|
|||
|
|
howpublished={\url{https://nehmeailabs.com}}
|
|||
|
|
}
|
|||
|
|
```
|
|||
|
|
|