Robustness Evaluation of GPU Cloud Pricing Benchmarks under Incomplete Observations: A Controlled Missingness Study

Impact 5 arXiv cs.AI

Summary

  • This study evaluates the robustness of daily GPU cloud rental-price benchmarks under incomplete marketplace observations.
  • The research stress-tests a fixed equal-weight median rule across four imposed masking mechanisms and four recovery methods using Monte Carlo simulations.
  • Whole-source outages reduce publication coverage to 0.77 while preserving zero conditional error on published dates.
  • Composition-changing data loss retains 0.86 coverage but generates 7.74 log points of error, whereas a past-only stratum-level ablation reduces error significantly.

Why it matters

Understanding the impact of incomplete pricing observations on GPU rental benchmarks helps cloud infrastructure planners and financial analysts build more resilient cost tracking models.

Who should care

finance 70