After consulting with experts in the topic of AI, I learned, that the accuracy of these models will suffer greatly, unless trained with the most accurate of datasets.
The LLM would have to be trained on 2 things:
Describing CVEs | Technical
Learning how to generate great headings | Marketing
The kind of natural language that would be required to generate the headings we are looking for, is both of technical and marketing nature.
A marketing expert knows how to create headings to attract a reader.
A technical expert knows which terms are necessary to deliver relevant information about a CVE.
The LLM would have to be a mix of both of these languages, which will require more effort, than it is worth.
We will look for a different alternative to describe CVEs, other than LLMs, in the future.
Using an LLM to describe CVEs will be an extensive task.
There are resources like this one, that allow training a model for describing CVEs. https://huggingface.co/datasets/morpheuslord/cve-llm-training
After consulting with experts in the topic of AI, I learned, that the accuracy of these models will suffer greatly, unless trained with the most accurate of datasets. The LLM would have to be trained on 2 things:
A marketing expert knows how to create headings to attract a reader. A technical expert knows which terms are necessary to deliver relevant information about a CVE. The LLM would have to be a mix of both of these languages, which will require more effort, than it is worth.
We will look for a different alternative to describe CVEs, other than LLMs, in the future.