Reconstruction of the SmCo Round 2 experiment-selection history

Reconstruct the SmCo Round 2 selection process by obtaining the unavailable Bayesian-optimization recommendations and mapping them to the executed wells.

Background

The saved SmCo Bayesian-optimization state records model and acquisition configuration but has an empty optimization history and no recommendation list. Consequently, the relationship between the recorded Round 2 wells and the selection policy cannot presently be established.

Recovering the recommendations and their well mapping would clarify how the experimental campaign was selected and would support a valid reconstruction of its decision process.

References

Reconstructing the Round~2 selection also requires the unavailable BO recommendations and their mapping to executed wells.

— Decision-Focused Active Learning for Scale-Aware Critical-Materials Recovery  (2609.09413 - Srinivas et al., 8 Sep 2026) in Section 7, “Related work, limitations, and open questions”