OpenRCA
Summary
OpenRCA is an LLM and agent benchmark for software root cause analysis. It asks models to reason over natural-language failure queries, KPI time series, dependency trace graphs, and semi-structured logs.
Official Artifacts
- OpenReview: https://openreview.net/forum?id=M4qNIzQYpd
- Official GitHub: https://github.com/microsoft/OpenRCA
- Dataset metadata snapshot: openrca-2025
Dataset Shape
The paper reports 335 failures from three enterprise software systems and more than 68 GB of telemetry. The GitHub README names the systems as Telecom, Bank, and Market.
Role In The Wiki
OpenRCA belongs in the LLM-agent RCA bucket, not the pure numeric graph-forecasting bucket. It is useful for testing tool-using telemetry investigation and exact root-cause reporting.
Relation To Foundation TSFM Agenda
Use the source-level agenda mapping in openrca-2025 rather than duplicating verdict rows here.
At the entity level, OpenRCA belongs in the LLM-agent RCA bucket, not the pure numeric graph-forecasting bucket. It is useful for testing tool-using telemetry investigation and exact root-cause reporting. This page should stay as the object card; source pages carry slot-level verdicts, evidence, and missing pieces.