tyaroshko/roberta-base-fake-news-detection

fake_news_model

This model is a fine-tuned version of FacebookAI/roberta-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0287
• Accuracy: 0.9949
• Precision: 0.9949
• Recall: 0.9949
• F1: 0.9949

Model description

More information needed

Intended uses & limitations

More information needed

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
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 4
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
0.6615	0.0718	100	0.5282	0.7945	0.8461	0.7945	0.7814
0.0469	0.1436	200	0.0299	0.9940	0.9940	0.9940	0.9940
0.0293	0.2154	300	0.0192	0.9962	0.9962	0.9962	0.9962
0.0344	0.2873	400	0.0141	0.9968	0.9968	0.9968	0.9968
0.0089	0.3591	500	0.0142	0.9971	0.9971	0.9971	0.9971
0.03	0.4309	600	0.0207	0.9962	0.9962	0.9962	0.9962
0.0074	0.5027	700	0.0354	0.9941	0.9941	0.9941	0.9941
0.0375	0.5745	800	0.0287	0.9949	0.9949	0.9949	0.9949

Framework versions

• Transformers 4.57.3
• Pytorch 2.9.1+cu128
• Datasets 4.4.1
• Tokenizers 0.22.1