MeCoLog: Meta-contrastive learning for cross-system few-shot log-based anomaly detection
Manh Tuan Nguyen, Tuan Phong Tran, Le Dinh Trang Dang, Van Loi Cao · 2026
Read MeCoLog: Meta-contrastive learning for cross-system few-shot log-based anomaly detection by Manh Tuan Nguyen, Tuan Phong Tran, Le Dinh Trang Dang, Van Loi Cao (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/plos-pone-0356907.txt.
Description
Manh Tuan Nguyen and colleagues built MeCoLog, a learning system that spots abnormal entries in computer system logs. It works even on a brand-new system with almost no labeled examples. Their 2026 paper tackles a practical headache. Every system writes logs differently. Detectors trained on one usually fail on another. New deployments have little history to learn from.
MeCoLog encodes each log entry by its type, value, and key parts. Unfamiliar wording still registers that way. It learns what normal behavior looks like across systems, while keeping anomalies strictly apart. Tested on five public log collections, it beat leading rival methods in the scarce-data cross-system setting. For operations teams, such transfer means a new service could gain decent monitoring from day one.
Catalog record
| CREATOR | Manh Tuan Nguyen, Tuan Phong Tran, Le Dinh Trang Dang, Van Loi Cao |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 68919 |
| LICENSE | CC_BY |
| RIGHTS | CC BY |
| SOURCE | PLOS |