AgentRCA: Zero-Shot Framework for Root Cause Analysis
AgentRCA, a novel framework, seeks to automate root cause analysis in industrial environments through a zero-shot agentic methodology. It integrates a data-driven digital twin that models standard system dynamics with a tool-enhanced large language model to enable reasoning during inference. The agent systematically collects statistical data, assesses alternative hypotheses, and pinpoints physical faults without the need for labeled examples of malfunctioning operations. This innovation overcomes the shortcomings of current black-box data-driven techniques, which struggle to provide justifications for diagnoses and rely on limited labeled data. The framework was detailed in a paper on arXiv (2607.22385) by an anonymous author.
Key facts
- AgentRCA is a zero-shot agentic framework for evidence-grounded root cause analysis.
- It combines a digital twin with a tool-augmented large language model.
- The agent iteratively gathers evidence, evaluates hypotheses, and identifies faults.
- It does not require labeled examples of faulty operation.
- The paper is available on arXiv with ID 2607.22385.
- Existing methods are black boxes and need scarce labeled data.
- The framework performs inference-time reasoning.
- It aims to automate the manual process of hypothesis formulation and evidence gathering.
Entities
Institutions
- arXiv