Even when RAG-based LLMs identify the right root-cause service 91–99% of the time, their recovery-action validity stays only 37–60% on a 302-incident Kubernetes benchmark.
Grace: A strategic llm-enhanced graph reinforcement learning framework for adaptive fault recovery in microservice systems,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SE 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Can LLMs Really Recover Microservice Failures? A Recovery-Aware Evaluation of Diagnosis-to-Action Reasoning
Even when RAG-based LLMs identify the right root-cause service 91–99% of the time, their recovery-action validity stays only 37–60% on a 302-incident Kubernetes benchmark.