Determine whether adding presolve rounds improves performance

Determine whether adding round boundaries beyond those present in SCIP’s default presolve policy can improve end-to-end mixed-integer programming solve time under the ORDO sequence-execution framework.

Background

ORDO permits round boundaries to be generated explicitly as sequence actions. The training corpus, however, is collected from the default SCIP policy, whose round structure is nearly unique; consequently, the corpus does not provide examples of sequences with additional rounds. The effect of expanding the round structure therefore remains unresolved by the reported experiments.

References

The claim of this paper therefore narrows to: under the round structure of the default policy, sequence-level decisions can significantly change end-to-end time; round boundaries are themselves generatable actions of this space, but the corpus contains no example of a round structure beyond the default policy's, so whether adding rounds can yield gains is not measured in this paper.

— ORDO: Operation-level Round-aware Dynamic Ordering for MIP Presolve  (2610.11294 - Chen et al., 8 Oct 2026) in Appendix B.12.2, Relation to the Configuration Space