Hybrid Risk-Aware Retrieval-Augmented Engineering Intelligence System for Automated Bug Analysis

Fuente: Zenodo
Salvato in:
Dettagli Bibliografici
Autore principale: dhruv, sharma
Natura: Recurso digital
Pubblicazione: Zenodo 2026
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866901156983209984
author dhruv, sharma
author_facet dhruv, sharma
contents <p>Software systems generate vast volumes of logs and operational data, making debugging increasingly complex and time-consuming. Developers and QA engineers often rely on manual inspection to identify failures, which is inefficient, error-prone, and difficult to scale in modern distributed systems.</p> <p>Recent advancements in <span><span>Large Language Models</span></span> (LLMs) have significantly improved capabilities in reasoning and code understanding. However, these models lack access to structured engineering data such as system logs, historical bug reports, and runtime signals, limiting their effectiveness in real-world debugging scenarios.</p> <p>To address this limitation, <span><span>Retrieval-Augmented Generation</span></span> (RAG) has been introduced to enhance LLM performance by incorporating external knowledge retrieval. While RAG improves contextual reasoning, it does not account for system-level risk signals that are critical for identifying and prioritizing high-impact failures.</p> <p>In this work, we propose a <strong>Hybrid Risk-Aware Retrieval-Augmented Engineering Intelligence System</strong>. The proposed system integrates machine learning-based risk prediction into the retrieval pipeline, enabling prioritization of high-risk failure patterns during the reasoning process. By incorporating system-level signals, the approach dynamically enhances the relevance and impact of retrieved context.</p> <p>Unlike traditional RAG-based debugging systems, our method introduces a <strong>risk-aware retrieval mechanism</strong> that prioritizes failure-critical information, resulting in more accurate, efficient, and actionable debugging insights.</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_19338786
institution Zenodo
language
publishDate 2026
publisher Zenodo
record_format zenodo
spellingShingle Hybrid Risk-Aware Retrieval-Augmented Engineering Intelligence System for Automated Bug Analysis
dhruv, sharma
<p>Software systems generate vast volumes of logs and operational data, making debugging increasingly complex and time-consuming. Developers and QA engineers often rely on manual inspection to identify failures, which is inefficient, error-prone, and difficult to scale in modern distributed systems.</p> <p>Recent advancements in <span><span>Large Language Models</span></span> (LLMs) have significantly improved capabilities in reasoning and code understanding. However, these models lack access to structured engineering data such as system logs, historical bug reports, and runtime signals, limiting their effectiveness in real-world debugging scenarios.</p> <p>To address this limitation, <span><span>Retrieval-Augmented Generation</span></span> (RAG) has been introduced to enhance LLM performance by incorporating external knowledge retrieval. While RAG improves contextual reasoning, it does not account for system-level risk signals that are critical for identifying and prioritizing high-impact failures.</p> <p>In this work, we propose a <strong>Hybrid Risk-Aware Retrieval-Augmented Engineering Intelligence System</strong>. The proposed system integrates machine learning-based risk prediction into the retrieval pipeline, enabling prioritization of high-risk failure patterns during the reasoning process. By incorporating system-level signals, the approach dynamically enhances the relevance and impact of retrieved context.</p> <p>Unlike traditional RAG-based debugging systems, our method introduces a <strong>risk-aware retrieval mechanism</strong> that prioritizes failure-critical information, resulting in more accurate, efficient, and actionable debugging insights.</p>
title Hybrid Risk-Aware Retrieval-Augmented Engineering Intelligence System for Automated Bug Analysis
url https://doi.org/10.5281/zenodo.19338786