---
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()

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: 2.0132
- Accuracy: 0.8191
- F1 Macro: 0.7488
- Low Precision: 0.6601
- Low Recall: 0.5006
- Low F1: 0.5694
- Medium Precision: 0.8440
- Medium Recall: 0.8767
- Medium F1: 0.8601
- High Precision: 0.8195
- High Recall: 0.8112
- High F1: 0.8153
- Critical Precision: 0.7618
- Critical Recall: 0.7392
- Critical F1: 0.7503

### 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 |
|:-------------:|:-----:|:-----:|:---------------:|:--------:|:--------:|:-------------:|:----------:|:------:|:----------------:|:-------------:|:---------:|:--------------:|:-----------:|:-------:|:------------------:|:---------------:|:-----------:|
| 2.3936        | 1.0   | 16180 | 2.5423          | 0.7404   | 0.6271   | 0.6925        | 0.2372     | 0.3534 | 0.7777           | 0.8359        | 0.8057    | 0.7237         | 0.7233      | 0.7235  | 0.6416             | 0.6110          | 0.6259      |
| 2.5847        | 2.0   | 32360 | 2.2926          | 0.7674   | 0.6790   | 0.6162        | 0.3880     | 0.4762 | 0.7899           | 0.8604        | 0.8237    | 0.7640         | 0.7458      | 0.7548  | 0.7115             | 0.6175          | 0.6612      |
| 2.0935        | 3.0   | 48540 | 2.1257          | 0.7920   | 0.7086   | 0.6727        | 0.4017     | 0.5030 | 0.8166           | 0.8670        | 0.8411    | 0.7907         | 0.7774      | 0.7840  | 0.7206             | 0.6927          | 0.7064      |
| 1.4077        | 4.0   | 64720 | 2.0427          | 0.8080   | 0.7367   | 0.5928        | 0.5203     | 0.5542 | 0.8334           | 0.8691        | 0.8509    | 0.8127         | 0.7952      | 0.8038  | 0.7583             | 0.7185          | 0.7379      |
| 1.0097        | 5.0   | 80900 | 2.0132          | 0.8191   | 0.7488   | 0.6601        | 0.5006     | 0.5694 | 0.8440           | 0.8767        | 0.8601    | 0.8195         | 0.8112      | 0.8153  | 0.7618             | 0.7392          | 0.7503      |


### Framework versions

- Transformers 5.6.2
- Pytorch 2.11.0+cu130
- Datasets 4.8.5
- Tokenizers 0.22.2