XLM_RoBERTa-Multilingual-Clickbait-Detection

This model is a fine-tuned version of xlm-roberta-large on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2192
• Micro F1: 0.9759
• Macro F1: 0.9758
• Accuracy: 0.9759

Test Set Macro-F1 scores

• Multilingual test set: 97.28
• en test set: 97.83
• el test set: 97.32
• it test set: 97.54
• es test set: 97.67
• ro test set: 97.40
• de test set: 97.40
• fr test set: 96.90
• pl test set: 96.18

Intended uses & limitations

• This model will be employed for an EU project.

Training and evaluation data

• The "clickbait_detection_dataset" was translated from English to Greek, Italian, Spanish, Romanian, French and German using the Opus-mt.
• The dataset was also translated from English to Polish using the M2M NMT.
• The "EasyNMT" library was utilized to employ the NMT models.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 4

Framework versions

• Transformers 4.36.1
• Pytorch 2.1.0+cu121
• Datasets 2.13.1
• Tokenizers 0.15.0