classify-clickbait-titles

Identify Clickbait Articles

This model is a fine-tuned version of albert/albert-base-v2 on a synthetic dataset with 65% factual article titles and 35% clickbait articles.

Built to demonstrate the use of synthetic data following, see the article here.

Model description

Built to identify factual vs clickbait titles.

Intended uses & limitations

Use it on any title to understand how the model is interpreting the title, whether it is factual or clickbait.

Go ahead and try a few of your own.

Here are a few examples:

Title: A Comprehensive Guide for Getting Started with Hugging Face Output: Factual

Title: OpenAI GPT-4o: The New Best AI Model in the World. Like in the Movies. For Free Output: Clickbait

Title: GPT4 Omni — So much more than just a voice assistant Output: Clickbait

Title: Building Vector Databases with FastAPI and ChromaDB Output: Factual

Training and evaluation data

It achieves the following results on the evaluation set:

• Loss: 0.0173
• Accuracy: 0.9951
• F1: 0.9951
• Precision: 0.9951
• Recall: 0.9951
• Accuracy Label Clickbait: 0.9866
• Accuracy Label Factual: 1.0

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 3

Framework versions

• Transformers 4.41.0
• Pytorch 2.2.1+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1