Publication
When Log Parsing Removes What Matters: Detecting Failures in Distributed GPU Workloads
Boyuan Cui / Yuyang Zeng / Yifang Tian / Hans-Arno Jacobsen
Proceedings of the 1st International Workshop on Reliable and trustworthy Automated Software Engineering (RASE '26), pp. 60–63, 2026 · Workshop Paper
Abstract
Read the abstract and full publication at the source.
Research outcomes
GF-Bench artifacts
Tools for generating multi-GPU failures, collecting NCCL logs, and evaluating anomaly detection, with sample data and scripts to reproduce the experiments.
Additional notes
The preprint linked above was revised after publication and differs from the RASE 2026 paper linked above.
The paper introduces GF-Bench for generating controlled faults in distributed GPU workloads and examines how log parsing affects failure detection.