A production-fidelity benchmark finds that LLM database-operation agents achieve at most 17.9% safe recovery versus 93.4% for human DBAs.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DB 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
DBA-Bench: A Production-Fidelity Benchmark for LLM-Based Database Operations Agents
A production-fidelity benchmark finds that LLM database-operation agents achieve at most 17.9% safe recovery versus 93.4% for human DBAs.