ModernBERT-base Open-Source Text Classification Model - Efficient and Practical through Fine-tuning on New Arxiv Datasets

Modernbert Base Is New Arxiv Dataset

Developed by davanstrien

A model fine-tuned on the arxiv-new-datasets dataset based on ModernBERT-base, primarily used for text classification tasks

Large Language Model

Transformers

Open Source License:Apache-2.0 #arXiv paper classification #High-accuracy BERT #Academic text processing

Downloads 10.40k

Release Time : 4/25/2025

Model Overview

This model is a fine-tuned version of ModernBERT-base on the arxiv-new-datasets dataset, suitable for academic paper classification tasks.

Model Features

High accuracy

Achieves 94.5% accuracy and 94.39% F1 score on the evaluation set

Optimized training

Uses cosine learning rate scheduler and label smoothing techniques to optimize the training process

Academic domain applicability

Trained on arXiv paper datasets, suitable for academic text processing

Model Capabilities

Text classification

Academic paper analysis

Natural language understanding

Use Cases

Academic research

Paper classification

Classify arXiv academic papers

94.5% accuracy

Academic text analysis

Analyze and classify academic paper content

🚀 Transformers Model

This project is a fine - tuned model based on the transformers library. It leverages the power of pre - trained models to achieve excellent performance on specific tasks.

🚀 Quick Start

This model is a fine - tuned version of [answerdotai/ModernBERT - base](https://huggingface.co/answerdotai/ModernBERT - base) on the [davanstrien/arxiv - new - datasets](https://huggingface.co/datasets/davanstrien/arxiv - new - datasets) dataset. It achieves the following results on the evaluation set:

Loss: 0.3220
Accuracy: 0.945
F1: 0.9439

📚 Documentation

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: 3e - 05
train_batch_size: 8
eval_batch_size: 64
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: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 30
label_smoothing_factor: 0.1

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.5181	1.0	300	0.4495	0.8333	0.8051
0.3804	2.0	600	0.3134	0.93	0.9268
0.3083	3.0	900	0.3407	0.9233	0.9192
0.2449	4.0	1200	0.3304	0.9367	0.9370
0.219	5.0	1500	0.3293	0.94	0.9377
0.2095	6.0	1800	0.3735	0.9283	0.9294
0.205	7.0	2100	0.3220	0.945	0.9439
0.2029	8.0	2400	0.3404	0.9367	0.9338
0.2	9.0	2700	0.3431	0.9333	0.9330
0.1989	10.0	3000	0.3286	0.9383	0.9377
0.1996	11.0	3300	0.3339	0.9383	0.9365
0.1986	12.0	3600	0.3295	0.9433	0.9419

Framework versions

Transformers 4.48.2
Pytorch 2.6.0+cu124
Datasets 3.2.0
Tokenizers 0.21.0

📄 License

This project is licensed under the Apache 2.0 license.

Property	Details
Library Name	transformers
Model Type	Fine - tuned version of answerdotai/ModernBERT - base
Training Data	davanstrien/arxiv - new - datasets
Metrics	accuracy, f1

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご