roberta-base-finetuned-allsides

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

• Loss: 0.2598
• Accuracy: 0.9293
• F1: 0.9298

Model description

This model can be used to:

• Predict political bias (left/center/right) from news headlines or article summaries.
• Analyze media bias and perform content audits.
• Enrich datasets for stance detection or political discourse analysis.

Intended uses & limitations

• The model reflects biases present in the AllSides dataset.
• Misclassification may occur, especially on ambiguous or neutral content.
• “Center” labels can be less well-defined and harder to learn.
• Model predictions do not account for context like author intent or publication history.

Training and evaluation data

More information needed

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: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 4

Training results

Framework versions

• Transformers 4.50.3
• Pytorch 2.6.0+cu124
• Datasets 3.5.0
• Tokenizers 0.21.1