Research

Paper

TESTING March 16, 2026

Fine-tuning RoBERTa for CVE-to-CWE Classification: A 125M Parameter Model Competitive with LLMs

Authors

Nikita Mosievskiy

Abstract

We present a fine-tuned RoBERTa-base classifier (125M parameters) for mapping Common Vulnerabilities and Exposures (CVE) descriptions to Common Weakness Enumeration (CWE) categories. We construct a large-scale training dataset of 234,770 CVE descriptions with AI-refined CWE labels using Claude Sonnet 4.6, and agreement-filtered evaluation sets where NVD and AI labels agree. On our held-out test set (27,780 samples, 205 CWE classes), the model achieves 87.4% top-1 accuracy and 60.7% Macro F1 -- a +15.5 percentage-point Macro F1 gain over a TF-IDF baseline that already reaches 84.9% top-1, demonstrating the model's advantage on rare weakness categories. On the external CTI-Bench benchmark (NeurIPS 2024), the model achieves 75.6% strict accuracy (95% CI: 72.8-78.2%) -- statistically indistinguishable from Cisco Foundation-Sec-8B-Reasoning (75.3%, 8B parameters) at 64x fewer parameters. We release the dataset, model, and training code.

Metadata

arXiv ID: 2603.14911
Provider: ARXIV
Primary Category: cs.CR
Published: 2026-03-16
Fetched: 2026-03-17 06:02

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.14911v1</id>\n    <title>Fine-tuning RoBERTa for CVE-to-CWE Classification: A 125M Parameter Model Competitive with LLMs</title>\n    <updated>2026-03-16T07:15:57Z</updated>\n    <link href='https://arxiv.org/abs/2603.14911v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.14911v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>We present a fine-tuned RoBERTa-base classifier (125M parameters) for mapping Common Vulnerabilities and Exposures (CVE) descriptions to Common Weakness Enumeration (CWE) categories. We construct a large-scale training dataset of 234,770 CVE descriptions with AI-refined CWE labels using Claude Sonnet 4.6, and agreement-filtered evaluation sets where NVD and AI labels agree. On our held-out test set (27,780 samples, 205 CWE classes), the model achieves 87.4% top-1 accuracy and 60.7% Macro F1 -- a +15.5 percentage-point Macro F1 gain over a TF-IDF baseline that already reaches 84.9% top-1, demonstrating the model's advantage on rare weakness categories. On the external CTI-Bench benchmark (NeurIPS 2024), the model achieves 75.6% strict accuracy (95% CI: 72.8-78.2%) -- statistically indistinguishable from Cisco Foundation-Sec-8B-Reasoning (75.3%, 8B parameters) at 64x fewer parameters. We release the dataset, model, and training code.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CR'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CL'/>\n    <published>2026-03-16T07:15:57Z</published>\n    <arxiv:comment>9 pages, 2 figures, 6 tables. Dataset: https://huggingface.co/datasets/xamxte/cve-to-cwe Model: https://huggingface.co/xamxte/cwe-classifier-roberta-base</arxiv:comment>\n    <arxiv:primary_category term='cs.CR'/>\n    <author>\n      <name>Nikita Mosievskiy</name>\n    </author>\n  </entry>"
}