Datasets:
[Organizer] Submission rules and FAQ (eligibility, channels, conventions)
This thread consolidates the organizer rules for OSC-Superconductor. It is the canonical reference and questions are answered here.
Discovery track. Materials outside the active set (active 1,034 of the 4,832 candidate pool), including new compositions and polytypes, are accepted as discovery proposals. They are queued for the organizer downfolding pipeline and enter the ranking only after canonical downfolding and verification.
Scored / ranking track. Only the active set is ranked directly. Non-active osc_ids are handled as discovery (rule 1).
Canonical pipeline. The canonical SCF and downfolding path may not be skipped; results must follow the canonical convention for a like-for-like comparison. organizer-equal QE submissions with a complete reproducibility package are reviewed for reproduction.
Channels. The official channel is this discussions board. Email is secondary. Submission status is posted in the relevant thread.
Conventions. n_band is the per-spin filling (n = 2f) under the corrected band-filling convention; solver_delta = 0.1667 for the active table retrieved 2026-10-21.
Gold / method contributions are reviewed in their own thread with the reproducibility package (see #59). Discovery proposals are posted as their own thread (see #60).
Title: Clarification: magnetic materials and the non-spin-polarized downfold
convention
Hello,
Thank you for the challenge and for the detailed FAQ in this thread. We have a
clarification
question about how magnetic materials are handled in the Discovery ranking, so that we
report
our results in the intended convention.
From the published pipeline (pipeline/downfold.py) and scoring_meta.json, the
canonical
downfold appears to use a non-spin-polarized SCF, a single eigenvalue set, and a DOS
spin
factor of 2; the active-set anchor table also contains only entries tagged
magnetic=No.
For materials tagged magnetic=Yes in candidates.csv, could you clarify:
- Is the canonical non-spin-polarized value the official score for a magnetic
material
as well, and is such a material eligible for the Discovery ranking? Or are magnetic
materials screened out before scoring? - If a material's spin-polarized ground state (FM or AFM) differs substantially from
the
non-spin-polarized solution, does that affect eligibility or verification? - If participants compute a spin-polarized check (e.g., FM/AFM energies and moments),
would you like it included in a Discovery proposal, or is only the canonical value
expected?
To make this concrete: in one case we find that the canonical non-spin-polarized value
is
well above the current leaderboard top, while the spin-polarized ground state gives an
index
close to zero. We want to report this in the form that is most useful for your
verification,
and to avoid submitting entries that fall outside the intended scope.
Thank you for the careful questions on magnetic materials. Direct answers.
The official/canonical score is the non-spin-polarized value, computed with the same fixed protocol for every material (nspin=1, DOS spin factor 2). Materials tagged magnetic=Yes are not in the scored active set (the active anchors are magnetic=No), so they are handled on the Discovery track, scored with the canonical non-spin value. They are not silently screened out; a Discovery proposal for a magnetic material is evaluated under the canonical convention.
A substantially different spin-polarized ground state (FM or AFM) does not change the canonical score, because the canonical metric is non-spin by definition. It does not by itself change eligibility or verification. A large spin-polarization effect is recorded as a physical caveat, not as a scoring change.
Please report both. The canonical non-spin value is what the pipeline scores and is required. A spin-polarized check (FM/AFM total energies and magnetic moments) is welcome as supplementary information and helps us flag where the non-spin convention and the physical ground state diverge.
To be transparent about the limitation: the canonical value is a fixed uniform-protocol number chosen for cross-material comparability, not a converged physical result, and for magnetic materials it can differ substantially from the spin-polarized ground state. That is why the scored active set is restricted to magnetic=No. Magnetic materials are therefore Discovery-only under the convention, with the spin-polarized check reported for physical context. A separate review of whether magnetic materials should be scored under a spin-polarized protocol is noted and will be decided separately; it would require re-deriving affected scores, so the current convention stands until then.