This challenge is a part of The Web Conference 2025.
Uncorrectable Errors (UEs) of Dynamic Random Access Memory (DRAM) have been identified as a major failure cause in data centers. The multi-bit UE failure of High Bandwidth Memory (HBM) highly threatens the availability and reliability of servers and entire computing clusters. Forecasting UEs before enacting preemptive maintenance measures has emerged as a viable strategy for diminishing server outages. Some machine-learning based solutions have also been proposed.
However, the UEs prediction presents several challenges: data noise and extreme imbalance as the UEs are exceedingly rare in memory events; heterogeneous data sources as the DRAMs in the field come from different manufacturing or architecture platforms; distribution shifts due to hardware aging; and latent factors due to the dynamic access mechanism.
We cure a real-world memory error dataset that contains both micro and bit information and present a two-stage challenge for more efficient and generalized event prediction solutions. We believe the competition will provide a breeding ground to foster discussions and further progress on several important research topics towards real-world ML applications.