Estimate full-pool discovery-to-fresh recurrence

Estimate the full-pool overlap between discovery-stage persistent selections and selections obtained in fresh executions of the one-call and two-agent resume-screening procedures.

Background

Fresh validation was restricted to discovery-selected persistent cases rather than the complete borderline application pool. The targeted fresh-run cohorts therefore provide information about recurrence among selected cases but cannot characterize how all applications transition from discovery runs to fresh executions.

A complete estimate would require re-executing the full evaluation pool, including cases that were not persistent during discovery, under the relevant screening procedures and comparing their discovery and fresh outcomes.

References

Because cases that were not persistent during discovery were not re-executed, we cannot estimate full-pool discovery-to-fresh overlap.

— When Hiring Becomes Agent-Mediated: Evaluating Access and Recurrence in Two-Agent Résumé Screening  (2609.19530 - Gao et al., 17 Sep 2026) in Limitations, subsection “Fresh validation”

Reusing validation requests for archive search and final selection can favor noisy high scores; we have not measured this effect on a fresh selection set or with repeated evaluations of the final candidates chosen by validation.

— Adaptive-GEPA: Make Your Harness Fit Heterogeneous Requests  (2609.38762 - Chen et al., 30 Sep 2026) in Section 1, Introduction; Section 6, Limitations; Appendix, Section 'Repeated validation and selection' (label app:validation-reuse)