Elamur Groupantsschool-vidtvorant
The groupant beside models/school-vidtvorant.groupant in the Mova repository: what the code computes, what it never does, how to go back from its outputs, its invariants, cost and checks.
Facts
Implementation material. A Mova code model describes this code at one source revision: partial models, finite checks and their stated scope. It is not a reviewed general exposition, and its labels are shown as written, never raised.
Components
| component | label | where |
|---|---|---|
| approximant | partial | Forward |
| ast | partial leaf projections | |
| composition | partial | Composition |
| cost | measured 2026 11 04 | Cost |
| descriptant | present | Descriptant |
| invariant | source derived | Invariant |
Obernants
The typed relations around the forward model: the behaviour complement (what the code never does), the reverse reconstruction (every input compatible with an output) and specialised views.
| id | type | label | where |
|---|---|---|---|
nonbehavior | behavior_complement | scoped model | nonbehavior |
compatible-evidence | reverse_reconstruction | scoped model | compatible-evidence |
Model text
The group of the student's vidtvorant: the reference observer (world/examples/student_probe.rs), the executable
vidtvorant (world/src/knowledge.rs.groupant/student-vidtvorant.groupant/rust_approximant.rs) and its runner
(tools/approx/examples/student_vidtvorant.rs). The subject they model is school-runtime (world/src/knowledge.rs),
whose groupant holds the vidtvorant as a child fragment.
Forward
probe(course) → traces (choice, understood, scores, answer ns) and stage times; repetitions → medians; runner(traces, lesson files) → fitted vidtvorant per fold → out-of-fold predictions → agreement, disagreement classes, time errors, own speed. The wiring is checked by two equalities: probe choices = CLI trace choices (three courses), and repetition choices identical (ten courses × five).
Composition
The probe's observation (choice, understood, ns) is exactly the runner's label set; the vidtvorant reads only lesson text and the shown options, never the student's internal frames, scores or anchors (those in the trace are used only to classify disagreements). Errors do not compose into a bound: choice agreement 67.24% says nothing about correctness on other courses or switches.
Invariant
Labels come from the student, not from the right answers; folds never share a lesson (parity) or a course (course split); semantic reports are deterministic; timing is reported apart.
Cost
Student: ≈ 101 s for ten courses on one core (median of five, under load); vidtvorant replay 0.48–0.56 s; fit and report ≈ 170 s.
nonbehavior
The group does not substitute the vidtvorant for the student anywhere, does not train Mova, call a teacher, read a held-out set, or claim a 1/2/5/10% cap (none is met, 2026-11-04).
compatible-evidence
The reports are compatible with any student revision that gives the same choices and similar times on the ten
courses; a new revision of knowledge.rs requires new traces (the fragment's source hash binds it).
Descriptant
2026-11-04: first epoch. Speed of the pupil and the fidelity of its word-bag shadow, measured on one P-core thread.
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/models/school-vidtvorant.groupant/