vijil-bias-detector-v1

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

• Loss: 0.3762
• Accuracy: 0.7799
• F1: 0.7449
• Precision: 0.8761
• Recall: 0.6479

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: 32
• eval_batch_size: 64
• 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
• lr_scheduler_warmup_steps: 0.1
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
0.5770	0.2082	500	0.5648	0.6727	0.6252	0.7236	0.5503
0.4443	0.4163	1000	0.4895	0.7199	0.7296	0.7000	0.7619
0.4140	0.6245	1500	0.4922	0.7519	0.7203	0.8173	0.6439
0.3954	0.8326	2000	0.3979	0.7693	0.7442	0.8269	0.6766
0.3955	1.0408	2500	0.3843	0.7735	0.7385	0.8640	0.6449
0.3780	1.2490	3000	0.3770	0.7834	0.7723	0.8068	0.7406
0.3884	1.4571	3500	0.3806	0.7758	0.7368	0.8820	0.6327
0.3749	1.6653	4000	0.3758	0.7725	0.7677	0.7777	0.7579
0.3957	1.8734	4500	0.3762	0.7799	0.7449	0.8761	0.6479

Framework versions

• Transformers 5.4.0
• Pytorch 2.11.0+cu130
• Datasets 4.8.4
• Tokenizers 0.22.2