Multi-Agent Recursion-of-Thought for Microservice Root Cause Localization
A new approach to root cause localization (RCL) in microservices uses multi-agent recursion-of-thought. Traditional machine learning and deep learning methods suffer from limited interpretability and poor transferability. Large language model (LLM)-based methods have been proposed but face context explosion and serial reasoning structures that degrade accuracy and efficiency. The paper studies how human site reliability engineers (SREs) perform RCL in practice, aiming to improve LLM-based methods. The work is published on arXiv with ID 2605.14866.
Key facts
- arXiv paper ID 2605.14866
- Focus on root cause localization for microservices
- Multi-agent recursion-of-thought approach
- Addresses limitations of traditional ML and DL methods
- LLM-based methods face context explosion and serial reasoning issues
- Studies human SRE practices
- Aims to improve interpretability and transferability
- Published on arXiv
Entities
Institutions
- arXiv