BGL
A log anomaly detection benchmark dataset derived from a supercomputer, commonly used to evaluate log anomaly detection methods.
Timeline (last 90 days)
LogCDM achieves an F1 score of 0.870 on the chronologically split BGL dataset.
On the BGL dataset, LogCDM outperforms all compact baselines.
Ablation experiments show that the structural signal brings F1 gains of 0.025-0.030 on BGL and Thunderbird, but contributes almost nothing on HDFS.
The evaluation protocol in the paper is a unified protocol, using a random split for HDFS and a chronological split for BGL.
How the chronologically split BGL result (0.870) can be fairly compared with conventionally reported random-split results still requires community consensus.
All Facts (5)
LogCDM achieves an F1 score of 0.870 on the chronologically split BGL dataset.
50%UnverifiedOn the BGL dataset, LogCDM outperforms all compact baselines.
50%UnverifiedAblation experiments show that the structural signal brings F1 gains of 0.025-0.030 on BGL and Thunderbird, but contributes almost nothing on HDFS.
50%UnverifiedThe evaluation protocol in the paper is a unified protocol, using a random split for HDFS and a chronological split for BGL.
50%UnverifiedHow the chronologically split BGL result (0.870) can be fairly compared with conventionally reported random-split results still requires community consensus.
50%