Qwen3-8B-UnBias-Plus-SFT-Instruct/README.md

---
license: apache-2.0
base_model: Qwen/Qwen3-8B
tags:
  - qwen3
  - bias-detection
  - news-debiasing
  - text-generation
  - fine-tuned
  - unsloth
  - sft
datasets:
  - vector-institute/Unbias-plus
language:
  - en
---
# Qwen3-8B-UnBias-Plus-SFT-Instruct

A fine-tuned version of [Qwen3-8B](https://huggingface.co/Qwen/Qwen3-8B) for news media bias detection and neutral rewriting, developed by the [Vector Institute](https://vectorinstitute.ai) as part of the [UnBias-Plus](https://github.com/VectorInstitute/unbias-plus) project.

Given a news article, the model identifies biased language segments, classifies their bias type and severity, provides neutral replacements, and returns a fully rewritten unbiased version of the article — all in a single structured JSON response.

This is the **Instruct variant** — trained without chain-of-thought thinking blocks (`enable_thinking=False`). It produces clean structured JSON directly, making it faster and more reliable for production inference, including deployment via vLLM or other OpenAI-compatible serving backends.

## Difference from [Qwen3-8B-UnBias-Plus-SFT](https://huggingface.co/vector-institute/Qwen3-8B-UnBias-Plus-SFT)

| | SFT (thinking) | SFT-Instruct (this model) |
|---|---|---|
| Thinking mode | `enable_thinking=True` | `enable_thinking=False` |
| Output | `<think>...</think>` + JSON | JSON directly |
| Inference backend | Transformers | Transformers / vLLM |
| Latency | Higher | Lower |
| Recommended for | Research, local use | Production APIs, vLLM deployment |

## Model Details

| Property | Value |
|---|---|
| Base model | Qwen/Qwen3-8B |
| Fine-tuning method | Supervised Fine-Tuning (SFT) with LoRA |
| Training precision | bf16 (full precision, no quantization during training) |
| LoRA rank | 16 |
| Training framework | Unsloth + TRL |
| Context length | 8192 tokens |
| Thinking mode | Disabled (`enable_thinking=False`) |
| Output format | Structured JSON |

## Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch, json

model_id = "vector-institute/Qwen3-8B-UnBias-Plus-SFT-Instruct"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    dtype=torch.bfloat16,
    device_map="auto",
)
model.eval()

SYSTEM_PROMPT = """You are an expert linguist and bias detection specialist.
Your task is to carefully read a news article, detect ALL biased language,
and return a structured JSON response. Return ONLY valid JSON, no extra text."""

article = "Your news article here..."

messages = [
    {"role": "system", "content": SYSTEM_PROMPT},
    {"role": "user", "content": f"Analyze the following article for bias and return the result in the required JSON format.\n\nARTICLE:\n{article}"},
]

inputs = tokenizer.apply_chat_template(
    messages,
    tokenize=True,
    add_generation_prompt=True,
    enable_thinking=False,   # must be False for this variant
    return_tensors="pt",
    return_dict=True,
    truncation=True,
    max_length=8192,
)

with torch.no_grad():
    outputs = model.generate(
        input_ids=inputs["input_ids"].to(model.device),
        attention_mask=inputs["attention_mask"].to(model.device),
        max_new_tokens=4096,
        do_sample=False,       # greedy decoding for deterministic JSON
        temperature=None,
        top_p=None,
        pad_token_id=tokenizer.eos_token_id,
    )

new_tokens = outputs[0][inputs["input_ids"].shape[1]:]
response = tokenizer.decode(new_tokens, skip_special_tokens=True)
result = json.loads(response)
```

### Using with the UnBias-Plus toolkit

```python
from unbias_plus import UnBiasPlus

pipe = UnBiasPlus(
    model_name_or_path="vector-institute/Qwen3-8B-UnBias-Plus-SFT-Instruct",
    load_in_4bit=False,   # set True for ~5GB VRAM
)

result = pipe.analyze("Your article text here...")
print(result.binary_label)      # "biased" or "unbiased"
print(result.severity)          # 0, 2, 3, or 4
print(len(result.biased_segments))
print(result.unbiased_text)
```

### Using with vLLM

```bash
vllm serve vector-institute/Qwen3-8B-UnBias-Plus-SFT-Instruct \
  --max-model-len 8192
```

```python
from openai import OpenAI

client = OpenAI(base_url="http://localhost:8000/v1", api_key="EMPTY")

completion = client.chat.completions.create(
    model="vector-institute/Qwen3-8B-UnBias-Plus-SFT-Instruct",
    messages=messages,
    max_tokens=4096,
    temperature=0,
    extra_body={"chat_template_kwargs": {"enable_thinking": False}},
)
result = json.loads(completion.choices[0].message.content)
```

## Output Schema

```json
{
  "binary_label": "biased" | "unbiased",
  "severity": 0 | 2 | 3 | 4,
  "bias_found": true | false,
  "biased_segments": [
    {
      "original": "exact substring from input article",
      "replacement": "neutral alternative phrase",
      "severity": "high" | "medium" | "low",
      "bias_type": "loaded language | dehumanizing framing | false generalizations | framing bias | euphemism/dysphemism | politically charged terminology | sensationalism",
      "reasoning": "1-2 sentence explanation"
    }
  ],
  "unbiased_text": "Full rewritten neutral article"
}
```

### Severity Scale

| Value | Meaning |
|---|---|
| 0 | Neutral — no bias detected |
| 2 | Recurring biased framing |
| 3 | Strong persuasive tone |
| 4 | Inflammatory rhetoric |

## Bias Types Detected

- **Loaded language** — words with strong emotional connotations
- **Dehumanizing framing** — language that strips dignity from groups
- **False generalizations** — sweeping statements ("they always", "all of them")
- **Framing bias** — selective wording that implies a viewpoint
- **Euphemism/dysphemism** — softening or hardening language to manipulate perception
- **Politically charged terminology** — labels used to provoke rather than describe
- **Sensationalism** — exaggerated language to evoke emotional responses

## Training Data

Fine-tuned on [vector-institute/Unbias-plus](https://huggingface.co/datasets/vector-institute/Unbias-plus), a curated dataset of news articles with expert-annotated bias labels, segment-level annotations, and neutral rewrites.

## Hardware Requirements

| Setup | Configuration |
|---|---|
| Recommended (server) | `load_in_4bit=False, dtype=torch.bfloat16` (~16GB VRAM) |
| Lightweight (laptop) | `load_in_4bit=True` (~5GB VRAM) |

## Model Variants

| | [4B SFT](https://huggingface.co/vector-institute/Qwen3-4B-UnBias-Plus-SFT) | [8B SFT](https://huggingface.co/vector-institute/Qwen3-8B-UnBias-Plus-SFT) | 8B SFT-Instruct (this) |
|---|---|---|---|
| VRAM (bf16) | ~8GB | ~16GB | ~16GB |
| VRAM (4-bit) | ~3GB | ~5GB | ~5GB |
| Thinking mode | ✓ | ✓ | ✗ |
| vLLM compatible | Partial | Partial | ✓ |
| Quality | Strong | Higher | Higher |
| Recommended for | Laptops | Research | Production APIs |

## Limitations

- Trained primarily on English-language news articles
- Political bias detection reflects patterns in the training data
- Best performance on articles under 5000 characters
- As with all language models, outputs should be reviewed by a human before use in production

## Citation

If you use this model in your research or application, please cite:

```bibtex
@misc{unbias-plus-8b-instruct,
  title        = {Qwen3-8B-UnBias-Plus-SFT-Instruct},
  author       = {Vector Institute},
  year         = {2026},
  publisher    = {HuggingFace},
  howpublished = {\url{https://huggingface.co/vector-institute/Qwen3-8B-UnBias-Plus-SFT-Instruct}},
  note         = {Part of the UnBias-Plus project: https://github.com/VectorInstitute/unbias-plus}
}
```

## Links

- 🔗 Project: [UnBias-Plus on GitHub](https://github.com/VectorInstitute/unbias-plus)
- 📊 Dataset: [vector-institute/Unbias-plus](https://huggingface.co/datasets/vector-institute/Unbias-plus)
- 🏛️ Organization: [Vector Institute](https://vectorinstitute.ai)
- 🤖 4B version: [vector-institute/Qwen3-4B-UnBias-Plus-SFT](https://huggingface.co/vector-institute/Qwen3-4B-UnBias-Plus-SFT)
- 🤖 8B thinking version: [vector-institute/Qwen3-8B-UnBias-Plus-SFT](https://huggingface.co/vector-institute/Qwen3-8B-UnBias-Plus-SFT)