A benchmark of **103 chaos-drill–injected incidents** across a Kubernetes + OpenTelemetry demo store, with per-task observability dumps and four-layer ground-truth annotations suitable for Root-Cause-Analysis (RCA) agent evaluation.
@@ -7,7 +7,13 @@ A benchmark of **103 chaos-drill–injected incidents** across a Kubernetes + Op
| **Tasks** | 103 (opaque IDs `t001`..`t103`) |
| **Cluster** | OpenTelemetry demo store on Alibaba Cloud ACK |
> **v1.1 release note (2026-07-01):** metric entity identifiers/names were backfilled, task alert entities were normalized, ReplicaSet topology labels were corrected, and internal load-test/build identifiers in traces were redacted as `<load-test-agent>` / `<internal-build>`. Task and row counts are unchanged from v1.0.
Public cases: `https://aiops-benchmark.oss-cn-hongkong.aliyuncs.com/rca/rca100/v1.1/`
-`mapping.json` — resolves `t001..t103` to canonical case IDs of the form `F<id>-<faultType>.<transId>`.
-`taxonomy.json` — defines the fault hierarchy and partial-credit scoring rubric.
Scoring: `Final = 0.4 · Entity + 0.3 · Fault + 0.3 · Process`, where Process jointly measures causal-chain node match rate and observability-checkpoint hit rate (numeric condition must match, not merely the metric name).