fakenews-detector-smollm2

Fake News Detector - SmolLM2 1.7B

An AI-powered misinformation detection model fine-tuned from SmolLM2-1.7B-Instruct on the Intel/misinformation-guard dataset.

What it does

Paste any news headline, social media claim, or statement and the model will:

1. Classify it as TRUE, MOSTLY TRUE, PARTIALLY TRUE, or FALSE
2. Explain its reasoning with evidence-based analysis

How to use

from transformers import pipeline

detector = pipeline("text-generation", model="pratsing23/fakenews-detector-smollm2")

messages = [
    {"role": "system", "content": "You are a misinformation detection AI. Analyze the given claim and determine if it is TRUE, MOSTLY TRUE, PARTIALLY TRUE, or FALSE. Provide clear reasoning."},
    {"role": "user", "content": 'Analyze this claim for misinformation: "5G towers cause COVID-19"'}
]

result = detector(messages, max_new_tokens=256, temperature=0.3)
print(result[0]["generated_text"][-1]["content"])

Output:

**Verdict: FALSE**

**Reasoning:** There is no scientific evidence linking 5G technology to COVID-19. 
The World Health Organization (WHO) has confirmed that viruses cannot travel on 
radio waves or mobile networks...

Performance

Strengths

• Perfect at confirming factual/real news (100%)
• Catches obvious misinformation (conspiracy theories, health hoaxes)
• Provides detailed reasoning for every verdict

Limitations

• Sometimes lenient with popular myths that have a "grain of truth"
• May label some false claims as "PARTIALLY TRUE" rather than "FALSE"

Training Details

Dataset Distribution

Use Cases

• Social media fact-checking - Verify viral claims
• News verification - Check headlines before sharing
• Educational tool - Learn to identify misinformation
• Content moderation - Flag potentially false content

Android App

This model powers the Misinfo Checker Android app. The app lets you paste any claim and get an instant AI-powered fact check on your phone.

License

Apache 2.0

Citation

@misc{fakenews-detector-smollm2,
  title={Fake News Detector SmolLM2},
  author={pratsing23},
  year={2026},
  publisher={HuggingFace},
  url={https://huggingface.co/pratsing23/fakenews-detector-smollm2}
}