RFR-2026-009: Cross-provider reproducibility semantics

Research opportunity

Define and test provider/model/role/run lineage so replication claims distinguish independence from repetition.

Background

Provider-specific effects are material and ROS currently lacks machine-resolved reproducibility semantics.

Origin documents and trace

• research/operating-system/evidence-dashboard/evidence-dashboard.json, section: current_claims; material_threats
• research/framework-engineering/ros-profile/open-ros-gaps.md, section: Cross-provider reproducibility semantics
• research/evaluations/FE-BOUNDARY-2026-07-28/ros-fe-profile-v1-audit.md, section: PA-015

Specific assumption challenged: Different provider labels alone imply independent replication.

Supporting evidence: The dashboard says provider differences remain material; the audit says dependencies and execution lineage are documentary.

Unknowns

• Operational degrees of independence
• Minimum lineage needed for replication
• Provider-by-instrument interaction

Dependencies

• RFR-2026-008

Suggested REP and methodology

Create REP-2026-009. Specify an independence taxonomy and lock record; validate it on repeated same-model, cross-model, cross-provider, and human runs.

Expected outputs

• Replication semantics standard
• Machine-readable lineage schema
• Validation cases

Success criteria

Consumers can mechanically distinguish rerun, robustness check, and independent replication.

Execution recommendation

• Recommended agent: AI reproducibility agent
• Estimated effort: medium
• Expected knowledge gained: Makes evidence-strength labels auditable.
• Frontier Score: 250
• Score inputs: knowledge gain 4/5; impact 4/5; cross-project reuse 4/5; scientific importance 4/5; dependency cost 3/5; implementation difficulty 3/5.
• Score calculation: 4 × 4 × 4 × 4 − 3 − 3 = 250.
• Score status: structured expert judgment, not measured utility.