End of training

README.md

@@ -1,67 +1,50 @@
 ---
 library_name: transformers
-license: cc-by-4.0
+license: mit
 base_model: roberta-base
-metrics:
-- accuracy
 tags:
 - generated_from_trainer
-- text-classification
-- classification
-- nlp
-- vulnerability
+metrics:
+- accuracy
 model-index:
 - name: vulnerability-severity-classification-roberta-base
   results: []
-datasets:
-- CIRCL/vulnerability-scores
 ---
 
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
 
-# 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.
+# vulnerability-severity-classification-roberta-base
 
+This model is a fine-tuned version of [roberta-base](https://huggingface.co/roberta-base) on an unknown dataset.
+It achieves the following results on the evaluation set:
+- Loss: 2.0502
+- Accuracy: 0.8139
+- F1 Macro: 0.7431
+- Low Precision: 0.6397
+- Low Recall: 0.4924
+- Low F1: 0.5565
+- Medium Precision: 0.8410
+- Medium Recall: 0.8714
+- Medium F1: 0.8559
+- High Precision: 0.8123
+- High Recall: 0.8011
+- High F1: 0.8067
+- Critical Precision: 0.7592
+- Critical Recall: 0.7477
+- Critical F1: 0.7534
 
 ## 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
+More information needed
 
-labels = ["low", "medium", "high", "critical"]
+## Intended uses & limitations
 
-model_name = "CIRCL/vulnerability-severity-classification-roberta-base"
-tokenizer = AutoTokenizer.from_pretrained(model_name)
-model = AutoModelForSequenceClassification.from_pretrained(model_name)
-model.eval()
+More information needed
 
-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)
+## Training and evaluation data
 
-# 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])
-```
+More information needed
 
 ## Training procedure
 
@@ -76,25 +59,20 @@ The following hyperparameters were used during training:
 - lr_scheduler_type: linear
 - num_epochs: 5
 
-It achieves the following results on the evaluation set:
-- Loss: 2.0151
-- Accuracy: 0.8167
-
-
 ### Training results
 
-| Training Loss | Epoch | Step  | Validation Loss | Accuracy |
-|:-------------:|:-----:|:-----:|:---------------:|:--------:|
-| 2.5823        | 1.0   | 15871 | 2.5155          | 0.7412   |
-| 2.2788        | 2.0   | 31742 | 2.2984          | 0.7668   |
-| 2.1224        | 3.0   | 47613 | 2.1535          | 0.7869   |
-| 1.8902        | 4.0   | 63484 | 2.0301          | 0.8057   |
-| 1.5144        | 5.0   | 79355 | 2.0151          | 0.8167   |
+| 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 |
+|:-------------:|:-----:|:-----:|:---------------:|:--------:|:--------:|:-------------:|:----------:|:------:|:----------------:|:-------------:|:---------:|:--------------:|:-----------:|:-------:|:------------------:|:---------------:|:-----------:|
+| 2.7509        | 1.0   | 15759 | 2.5739          | 0.7341   | 0.6469   | 0.5294        | 0.3716     | 0.4367 | 0.7947           | 0.8050        | 0.7998    | 0.7148         | 0.7127      | 0.7137  | 0.6192             | 0.6570          | 0.6375      |
+| 2.3507        | 2.0   | 31518 | 2.4020          | 0.7609   | 0.6754   | 0.5901        | 0.3763     | 0.4595 | 0.7847           | 0.8626        | 0.8218    | 0.7920         | 0.6852      | 0.7348  | 0.6399             | 0.7385          | 0.6857      |
+| 1.6387        | 3.0   | 47277 | 2.1924          | 0.7858   | 0.7093   | 0.6319        | 0.4335     | 0.5142 | 0.8197           | 0.8558        | 0.8374    | 0.7971         | 0.7467      | 0.7711  | 0.6734             | 0.7614          | 0.7147      |
+| 1.2756        | 4.0   | 63036 | 2.0808          | 0.8042   | 0.7289   | 0.6554        | 0.4562     | 0.5380 | 0.8217           | 0.8792        | 0.8495    | 0.8022         | 0.7858      | 0.7939  | 0.7730             | 0.6991          | 0.7342      |
+| 1.5348        | 5.0   | 78795 | 2.0502          | 0.8139   | 0.7431   | 0.6397        | 0.4924     | 0.5565 | 0.8410           | 0.8714        | 0.8559    | 0.8123         | 0.8011      | 0.8067  | 0.7592             | 0.7477          | 0.7534      |
 
 
 ### Framework versions
 
-- Transformers 5.4.0
+- Transformers 5.5.0
 - Pytorch 2.11.0+cu130
 - Datasets 4.8.4
 - Tokenizers 0.22.2

config.json

@@ -34,7 +34,7 @@
   "pad_token_id": 1,
   "problem_type": "single_label_classification",
   "tie_word_embeddings": true,
-  "transformers_version": "5.4.0",
+  "transformers_version": "5.5.0",
   "type_vocab_size": 1,
   "use_cache": true,
   "vocab_size": 50265

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:673e35db5e1847d0950d3f200fc2eea3e18bb4016e48ba81e15e4f0e4272ca7a
+oid sha256:45b332e5b806ab9f5eec0cc477a61ecc9319048bf8f539587b597f37388a418d
 size 498618952

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d5c9228ce48c1d74ec90ce865a9b5c409505e7e08212aacac7217e118b754c76
+oid sha256:6a89f00003aad1fd8409cc273592dc9478b1934fab4e9a8a3b21a4719f2d2dc3
 size 5265