Frequently asked questions

What does UNIQ²Me help me do?

It records which defined assessment conditions are satisfied and checks the subset relevant to a decision. Each answer remains tied to its question and supporting record. Start with the Introduction for the method and About for the project context.

What are an SF and an SFS?

A Scoring Factor (SF) is a binary question with explanations, an evidence criterion, references, and assessment metadata. A Scoring Factor Set (SFS) contains sixteen SFs. A catalog is the complete 16 × 16 questionnaire, with sixteen sets and 256 questions. A mask selects requirements within that catalog. Set and question indices run from 0 to F; see Scoring Factor Sets.

Is a larger hexadecimal profile better?

The value encodes answers in fixed positions; it is not a total score or a risk estimate. For example, 8000 has one satisfied question, while 7000 has three different satisfied questions. Which answers matter depends on the selected requirements. Scoring explains the encoding, and Qualitative Measurements explains how to interpret it.

How does a requirement mask work?

A 1 in the mask selects an SF that must be satisfied. A 0 excludes it from that check; it does not require a negative answer. The benchmark passes when (S AND M) = M, and (S AND M) XOR M identifies unmet selected requirements. See Requirement masks.

Which MASKS section should I use?

Use catalog masks with a complete sixteen-field profile, SFS masks with one four-digit set value, or individual-SF masks with resolved answers at explicitly selected positions. All three refer to a complete 16 × 16 catalog. An SFS or individual-SF check does not become a complete assessment. See terms and notation.

Can I edit and promote a mask?

Keep one JSON per mask in uniq2me/repo-sfs-masks/<language>/, starting from the supplied templates. Record its purpose, scope, evidence period, selection, references and catalog binding. A reviewed mask requires a published catalog edition, named maintainer and accepted, named and dated review. The current examples remain drafts. See mask authoring and review.

Why is a mask unavailable after a catalog change or translation?

The library checks its bound catalog digest against the resolved catalog. English fallback preserves the original binding; matching positions do not prove that translated questions mean the same thing. Review the mapping and create or revise a mask against the intended snapshot. See versions and bindings.

Does missing evidence mean a failed question?

Missing, contradictory, or outdated evidence needs clarification under the SF's assessment rule. Pending findings stay outside the binary encoding until resolved. An unfilled authoring field such as resultScore: null is not an assessed failure. The supporting record must explain partial fulfillment and applicability; see Validation of Scoring Factors.

Who decides the final answer?

The method uses one assessor or three assessors, retaining their judgments and reasons. With three ordinary judgments, the majority decides. Marked judgments such as 0+ and 1+ receive priority under the specified resolution rules; the plus sign is not statistical confidence. Only the resolved 0 or 1 enters the profile. See Validation of Scoring Factors for the rules.

Does passing a benchmark establish compliance or security?

It establishes that the resolved answers satisfy the selected mask under the recorded catalog and scope. It does not establish legal compliance, estimate risk, or guarantee operational effectiveness. The recovery example shows exactly what a small NIS2-related mask checks and where that conclusion ends.

What must match before I compare profiles?

Check the question definitions, bit ordering, catalog version, system scope, and evidence period. A changed question or scope may require an explicit mapping. Translations also need reviewed equivalence; matching identifiers alone do not guarantee matching meanings. See Versioning.

What should I keep alongside a profile?

Keep the exact catalog identifier, selected mask, system boundary, evidence period, evidence references, assessor judgments, and reasons for the resolved answers. These records let a reader reconstruct the finding. See question records for the proposed authoring design and assessment and validation for the distinction between a question definition and an assessment instance.

The browser comparator checks manually entered answers against a catalog-bound mask. It does not verify that those answers came from the stated system, evidence period or assessment record.

Can an open-source reference solution prove that an SF is satisfied?

A canonical solution gives a selected implementation approach and verification guidance. Alternative implementations can also satisfy the same criterion. Installing the named tool or following a recipe still requires evidence from the assessed system. LLM Training and Assessment Assistance explains how guidance and observed evidence serve different roles.

Can agents or language models perform the assessment?

The proposed workflow lets models interpret evidence, suggest judgments, and retrieve remediation guidance. Review resolves pending findings and accepts judgments before deterministic software encodes them. XOR can locate disagreements between agents, but accuracy requires an independent reference. See Use Cases and Examples and the LLM workflow.

Is the planned catalog or small language model available?

The manuscript describes a prepared complete catalog and canonical guidance, to be published with the finished paper alongside compatible reviewed MASKS. It uses the illustrative designation groupa/en/26a. The reviewed training corpus, model adaptation, and evaluation remain ongoing work; this documentation reports no available trained artifact or measured model gains. See LLM Training and Assessment Assistance for release details still to be supplied, Versioning for identifiers, and Resources for supporting material.