ModernBERT-Turkish-FakeNews-Classifier

This model is a fine-tuned version of artiwise-ai/modernbert-base-tr-uncased on a Fake News Detection dataset, which is constructed by combining various publicly available datasets, incorporating additional resources curated by me, and translating English-language data into Turkish using large language models. It achieves the following results on the evaluation set:

• Loss: 0.2734
• Accuracy: 0.9514
• Precision: 0.9180
• Recall: 0.9825
• F1: 0.9492

Training and evaluation data

The data cannot be shared.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• 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
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
0.2843	1.0	124	0.2695	0.9150	0.8780	0.9474	0.9114
0.1242	2.0	248	0.2612	0.9514	0.9113	0.9912	0.9496
0.009	3.0	372	0.2455	0.9514	0.9180	0.9825	0.9492
0.0017	4.0	496	0.2675	0.9514	0.9180	0.9825	0.9492
0.0	5.0	620	0.2734	0.9514	0.9180	0.9825	0.9492

Framework versions

• Transformers 4.52.2
• Pytorch 2.6.0+cu124
• Datasets 2.14.4
• Tokenizers 0.21.1