• qa_deck_100#slide 10 · Evidence pack and score profile

The run-level evidence pack and the score profile do different jobs inside RAIDT. The run-level evidence pack is the structured proof object: it records what happened in the run, including prompts, versions, retrieved material, outputs, checks, and provenance. On its own, however, the pack is descriptive rather than evaluative. It can show that evidence exists, but it does not by itself say whether that evidence is sufficient for governance. The score profile adds that second layer by applying a predefined rubric to the same artefacts across the five pillars (Responsibility, Auditability, Interpretability, Dependability, Traceability). In other words, the evidence pack is the source of truth, while the score profile states what the recorded evidence means for governance readiness.

This added layer matters because RAIDT is designed to move organisations away from narrative assurance and towards comparable judgement. The score profile makes reviewer expectations explicit through anchors 1=missing / 3=partial / 5=audit-ready, so the organisation can distinguish between a merely logged run and a genuinely reviewable one. It also makes trade-offs visible: a run can be well documented but still weak on Interpretability or Dependability. Once scores are attached, organisations can calibrate reviewers, compare configurations, set escalation thresholds, sample runs for audit, and target remediation where scores are persistently low. The profile therefore adds actionability, comparability, and governance learning beyond the raw existence of records. It does not replace the evidence pack; it converts that pack into a disciplined judgement about whether the run is fit for contestability, oversight, and improvement.

An HR team may have a run-level evidence pack for a shortlist justification workflow: the prompt template, job criteria, model deployment ID, and output are all stored. At first glance, that looks well governed. Yet if the adapter version is not recorded, the organisation cannot later reconstruct which PEFT or LoRA configuration shaped the output. The evidence pack exists, but the score profile would lower Traceability because provenance is incomplete.

That extra judgement is what the score profile adds. It tells the team that documentation is present but not sufficient. The remedial action is also clearer: version the adapter, record it in every run, and rescore. Without the score profile, the organisation might mistake record-keeping for governance readiness.

• 00-RAIDT_Wording_v2
• 08-RAIDT_Foundations_M_V50
• 11-RAIDT_Academic_Logic_M_v11