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
