Incremental predictive value of cumulative state load

Determine whether cumulative state load provides predictive value for model execution-prediction errors beyond the contributions of trace length and peak state size.

Background

The paper defines cumulative state load as the sum, over execution steps, of the reachable program-state size. Because this quantity combines execution length and state width, the authors examine whether it is associated with prediction failure. Their adjusted logistic analysis finds no positive association: the odds ratio per doubling is 0.859, with a 95% interval of [0.733, 1.006]. However, trace length, peak state size, and cumulative state load are strongly correlated, and the exploratory analysis does not resolve whether the composite measure contributes predictive information beyond its components.

References

Trace length, peak state size, and cumulative state load are correlated, and the present analysis does not establish whether the composite adds predictive value beyond its components.

— Evaluating Exact Output and Checkpoint-State Prediction in Real Programs  (2610.11889 - Chen et al., 8 Oct 2026) in Appendix, Section “Original Exploratory Metric Analyses” (following Figure stateload discussion)