You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: README.md
+1-1Lines changed: 1 addition & 1 deletion
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -30,7 +30,7 @@ Developing and evaluating agentic root cause analysis methods for network troubl
30
30
|---|---|---|
31
31
|**Vanishing faults**| Real network incidents resolve too fast to study or replay. Researchers can only chase snapshots after the fact. | Containerlab + SONiC-VS reproduce faults deterministically. Every run is an identical, repeatable episode. |
32
32
|**Static data**| Existing datasets are frozen logs or hand-crafted prompts — no live telemetry, no interactive tool calls, no true cause-and-effect chain. | Agents operate inside a running network. Pingmesh, gNMI, sFlow, and syslog evidence is generated in real time during each episode. |
33
-
|**No fair arena**| Different teams evaluate on different topologies, different faults, and different metrics. Comparisons are meaningless. | A shared scenario format, ground-truth labels, and a localization-first scorer make cross-agent and cross-model comparisons directly comparable. |
33
+
|**No fair comparison**| Different teams evaluate on different topologies, different faults, and different metrics. Comparisons are meaningless. | A shared scenario format, ground-truth labels, and a localization-first scorer make cross-agent and cross-model comparisons directly comparable. |
0 commit comments