Fine-Tuned MS MARCO MiniLM Cross-Encoder

This model is a fine-tuned version of cross-encoder/ms-marco-MiniLM-L6-v2.

Description

The model was further trained for cybersecurity-related passage ranking, question answering, and relevance scoring tasks.

Given a query and a document (or passage), the model produces a relevance score that can be used for reranking search results and retrieving the most relevant cybersecurity information.

Usage

from sentence_transformers import CrossEncoder

model = CrossEncoder("jobby32/ms-marco-cybersecurity-MiniLM-L6-v2")

score = model.predict([
    (
        "What is a SQL injection attack?",
        "SQL injection is a code injection technique used to exploit vulnerabilities in database-driven applications."
    )
])

print(score)

Base Model

• cross-encoder/ms-marco-MiniLM-L6-v2

Training Data

This model was fine-tuned on the jobby32/cybersecurity-QA-with-negatives dataset.

The dataset was created from publicly available cybersecurity question-answering datasets, including:

• Rowden/CybersecurityQAA
• sambanovasystems/attackqa
• mariiazhiv/cybersecurity_qa

Acknowledgements

Special thanks to the authors and maintainers of the original datasets for making their work publicly available to the cybersecurity and machine learning communities.

Intended Use

This model is intended for:

• Cybersecurity search and retrieval systems
• Document reranking
• Retrieval-Augmented Generation (RAG)
• Security knowledge bases
• Question answering pipelines

License

This model is based on cross-encoder/ms-marco-MiniLM-L6-v2 and follows the Apache-2.0 license.