Disciplines · Proposals

YSD-0130 — Gold-set and seeded-corpus licensing posture

What licensing posture governs evaluation gold sets and seeded corpora?

5sections1 minread

On this page
  • Status: approved (2026-07-18)
  • Priority: P0 EXT — blocking (production rollout gated via YSD-0139)
  • Decision owner: @GreyChimp
  • Drafted: 2026-07-18 by Claude Code (claude-fable-5)
  • Approval: approved as recommended by @GreyChimp on 2026-07-18; outcome recorded in decision-log.json; review by 2027-07-18

Question#

What licensing posture governs evaluation gold sets and seeded corpora?

Recommendation#

Three admitted source classes only: (1) public-domain/openly-licensed works with verified jurisdiction and license version (YSD-4008); (2) creator-owned/internal material with participant consent; (3) explicitly licensed evaluation sets with written grants covering annotation derivatives, evaluation use, and expert review. No scraped commercial media; no in-repo fabricated benchmark numbers (per the 2026-07-05 Meshy audit precedent). Expert-review terms drafted per engagement and recorded with the set.

Options considered#

  • Fair-use-based commercial sampling — rejected: region-dependent, revocable, and contradicts the deny-by-default rights posture.

Consequences#

  • Section 18.1 evaluation items consume only sets admitted under this posture; each gold set records its class, grant evidence, and permitted uses.

Machine-enforced outcome (once approved)#

Evaluation harness refuses sets without a recorded rights class + grant reference; YSD-18001+ items enforce.