---
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 = outputs.logits
# 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: 2.0079
- Accuracy: 0.8187
- F1 Macro: 0.7495
- Low Precision: 0.6490
- Low Recall: 0.5059
- Low F1: 0.5686
- Medium Precision: 0.8468
- Medium Recall: 0.8712
- Medium F1: 0.8588
- High Precision: 0.8140
- High Recall: 0.8114
- High F1: 0.8127
- Critical Precision: 0.7671
- Critical Recall: 0.7488
- Critical F1: 0.7579

### 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 |
|:-------------:|:-----:|:-----:|:---------------:|:--------:|:--------:|:-------------:|:----------:|:------:|:----------------:|:-------------:|:---------:|:--------------:|:-----------:|:-------:|:------------------:|:---------------:|:-----------:|
| 2.8382        | 1.0   | 16475 | 2.5695          | 0.7351   | 0.6552   | 0.4964        | 0.4244     | 0.4576 | 0.7986           | 0.7994        | 0.7990    | 0.7393         | 0.6933      | 0.7156  | 0.5868             | 0.7245          | 0.6484      |
| 2.3037        | 2.0   | 32950 | 2.3201          | 0.7709   | 0.6774   | 0.6294        | 0.3490     | 0.4490 | 0.8025           | 0.8541        | 0.8275    | 0.7629         | 0.7502      | 0.7565  | 0.6908             | 0.6630          | 0.6766      |
| 2.1765        | 3.0   | 49425 | 2.1006          | 0.7905   | 0.7077   | 0.6790        | 0.3867     | 0.4928 | 0.8246           | 0.8568        | 0.8404    | 0.7894         | 0.7659      | 0.7775  | 0.6903             | 0.7524          | 0.7201      |
| 1.7249        | 4.0   | 65900 | 2.0247          | 0.8091   | 0.7329   | 0.6677        | 0.4528     | 0.5396 | 0.8236           | 0.8874        | 0.8543    | 0.8136         | 0.7828      | 0.7979  | 0.7669             | 0.7144          | 0.7397      |
| 1.3227        | 5.0   | 82375 | 2.0079          | 0.8187   | 0.7495   | 0.6490        | 0.5059     | 0.5686 | 0.8468           | 0.8712        | 0.8588    | 0.8140         | 0.8114      | 0.8127  | 0.7671             | 0.7488          | 0.7579      |


### Framework versions

- Transformers 5.8.1
- Pytorch 2.12.0+cu130
- Datasets 4.8.5
- Tokenizers 0.22.2