Cyber Threat Intelligence reports combine analytical prose with dense technical indicators, making structured entity extraction a challenging but operationally valuable task. This paper presents a comparative evaluation of three large language models - Claude Sonnet 4.6, GPT-5.4, and LLaMA 4 Scout - on a manually annotated corpus of 21 real-world CTI reports across 15 entity types and 1284 ground truth instances. The paper evaluates zero-shot and few-shot prompting conditions and studies the effect of iterative prompt refinement, focusing on explicit format constraints for cryptographic hash entities. Results show that Claude Sonnet 4.6 and GPT-5.4 achieve comparable performance under zero-shot conditions, with LLaMA 4 Scout trailing by a substantial margin. Few-shot prompting consistently reduces hallucination rates but yields mixed F1 results, with exemplar cardinality emerging as a critical design factor. Entity extraction difficulty varies substantially across types, with technical indicator categories showing near-perfect performance and semantic categories such as tool and target sector posing the greatest challenges across all evaluated models.
| Mualliflar | Гусейнли, А. |
|---|---|
| Jurnal | Рақамли технологияларнинг назарий ва амалий масалалари |
| Nashr sanasi | 2026-08-03 |
| Jild | 9 |
| Son | 3 |
| Betlar | 7-15 |
| Til | Rus |
| DOI | 10.62132/ijdt.v9i3.393 |
DOI: 10.62132/ijdt.v9i3.393 · Maqolaning asl sahifasi
киберразведка, распознавание именованных сущностей, большие языковые модели, инженерия промптов, cyber threat intelligence, named entity recognition, large language models, prompt engineering
In a multi-agent information environment, facts are stored in local repositories maintained by individual agents. As repositories expand and the number of agents increases, evaluating every query against all facts and…
A coupled magneto-mechanical model to simulate the deformation of a ferromagnetic cylindrical shell under the action of an axial alternating magnetic field is presented. The model includes the anhysteretic Langevin…
This article analyzes the medical diagnosis process from a data perspective and describes an approach to formally representing a mixed-type feature space describing a patient's condition. The formation of medical data…
This paper proposes BCA-YOLO, a YOLO-based architecture enhanced with three novel components. Bilateral Cross-Attention (BCA) block, which computes cross-attention between the feature maps of the left breast and the…
Cardiovascular diseases are one of the leading causes of death worldwide, and their early detection is one of the urgent tasks of modern medicine. In recent years, machine learning algorithms have been widely used to…
This study is aimed at assessing the impact of groundwater dynamics on the noosphere using modern information systems and integrated modeling approaches. The research proposes the use of remote sensing data and GIS…
Phishing remains one of the most persistent cybersecurity threats, and URLs are often the earliest observable indicator of an attack. Although existing phishing URL detection methods achieve high accuracy using lexical…
This article analyzes the scientific-methodological foundations, architecture and pedagogical effectiveness of the «Mantiq» mobile platform, which monitors speech and cognitive development in preschool children using…
This paper presents a comprehensive review of existing methods and approaches for Sound Event Detection (SED) in environmental acoustic monitoring. The theoretical foundations of SED, acoustic feature extraction…
This article examines the geometric modeling of Islamic decorative patterns and digital algorithms for their creation. The compositional structure of Islamic patterns, the laws of symmetry, repeating elements, and the…
Рақамли технологияларнинг назарий ва амалий масалалари — barcha maqolalar