---
library_name: transformers
license: cc-by-4.0
base_model: roberta-base
metrics:
- accuracy
tags:
- generated_from_trainer
- text-classification
- classification
- nlp
- vulnerability
model-index:
- name: vulnerability-severity-classification-roberta-base
  results: []
datasets:
- CIRCL/vulnerability-scores
---


# VLAI: A RoBERTa-Based Model for Automated Vulnerability Severity Classification

# Severity classification

This model is a fine-tuned version of [roberta-base](https://huggingface.co/roberta-base) on the dataset [CIRCL/vulnerability-scores](https://huggingface.co/datasets/CIRCL/vulnerability-scores).

The model was presented in the paper [VLAI: A RoBERTa-Based Model for Automated Vulnerability Severity Classification](https://huggingface.co/papers/2507.03607) [[arXiv](https://arxiv.org/abs/2507.03607)].

**Abstract:** VLAI is a transformer-based model that predicts software vulnerability severity levels directly from text descriptions. Built on RoBERTa, VLAI is fine-tuned on over 600,000 real-world vulnerabilities and achieves over 82% accuracy in predicting severity categories, enabling faster and more consistent triage ahead of manual CVSS scoring. The model and dataset are open-source and integrated into the Vulnerability-Lookup service.

You can read [this page](https://www.vulnerability-lookup.org/user-manual/ai/) for more information.


## Model description

It is a classification model and is aimed to assist in classifying vulnerabilities by severity based on their descriptions.

## How to get started with the model

```python
from transformers import AutoModelForSequenceClassification, AutoTokenizer
import torch

labels = ["low", "medium", "high", "critical"]

model_name = "CIRCL/vulnerability-severity-classification-roberta-base"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)
model.eval()

print("Model revision:", model.config._commit_hash)

test_description = "SAP NetWeaver Visual Composer Metadata Uploader is not protected with a proper authorization, allowing unauthenticated agent to upload potentially malicious executable binaries \
that could severely harm the host system. This could significantly affect the confidentiality, integrity, and availability of the targeted system."
inputs = tokenizer(test_description, return_tensors="pt", truncation=True, padding=True)

# Run inference
with torch.no_grad():
    outputs = model(**inputs)
    predictions = torch.nn.functional.softmax(outputs.logits, dim=-1)

# Print results
print("Predictions:", predictions)
predicted_class = torch.argmax(predictions, dim=-1).item()
print("Predicted severity:", labels[predicted_class])
```

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 3e-05
- train_batch_size: 32
- eval_batch_size: 32
- 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

It achieves the following results on the evaluation set:
- Loss: 2.0411
- Accuracy: 0.8164
- F1 Macro: 0.7491
- Low Precision: 0.6536
- Low Recall: 0.5148
- Low F1: 0.5759
- Medium Precision: 0.8377
- Medium Recall: 0.8712
- Medium F1: 0.8541
- High Precision: 0.8236
- High Recall: 0.8077
- High F1: 0.8156
- Critical Precision: 0.7561
- Critical Recall: 0.7457
- Critical F1: 0.7509

### Training results

| Training Loss | Epoch | Step  | Validation Loss | Accuracy | F1 Macro | Low Precision | Low Recall | Low F1 | Medium Precision | Medium Recall | Medium F1 | High Precision | High Recall | High F1 | Critical Precision | Critical Recall | Critical F1 |
|:-------------:|:-----:|:-----:|:---------------:|:--------:|:--------:|:-------------:|:----------:|:------:|:----------------:|:-------------:|:---------:|:--------------:|:-----------:|:-------:|:------------------:|:---------------:|:-----------:|
| 2.6595        | 1.0   | 18151 | 2.5896          | 0.7341   | 0.6285   | 0.6716        | 0.2567     | 0.3714 | 0.7917           | 0.7991        | 0.7954    | 0.7002         | 0.7492      | 0.7239  | 0.6326             | 0.6141          | 0.6232      |
| 2.4978        | 2.0   | 36302 | 2.3990          | 0.7614   | 0.6814   | 0.5762        | 0.4255     | 0.4895 | 0.7779           | 0.8556        | 0.8149    | 0.7881         | 0.7133      | 0.7488  | 0.6681             | 0.6771          | 0.6726      |
| 1.7617        | 3.0   | 54453 | 2.1975          | 0.7879   | 0.7097   | 0.6682        | 0.4156     | 0.5125 | 0.8119           | 0.8615        | 0.8360    | 0.8056         | 0.7546      | 0.7793  | 0.6775             | 0.7484          | 0.7112      |
| 1.7304        | 4.0   | 72604 | 2.0677          | 0.8075   | 0.7366   | 0.6663        | 0.4761     | 0.5554 | 0.8389           | 0.8588        | 0.8487    | 0.8077         | 0.8021      | 0.8049  | 0.7248             | 0.7506          | 0.7375      |
| 1.4189        | 5.0   | 90755 | 2.0411          | 0.8164   | 0.7491   | 0.6536        | 0.5148     | 0.5759 | 0.8377           | 0.8712        | 0.8541    | 0.8236         | 0.8077      | 0.8156  | 0.7561             | 0.7457          | 0.7509      |


### Framework versions

- Transformers 5.15.0
- Pytorch 2.13.0+cu130
- Datasets 4.8.5
- Tokenizers 0.22.2