Elamur Groupantsschool-student-vidtvorant-runner
The groupant beside tools/approx/examples/student_vidtvorant.rs 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 | source derived | forward |
| ast | partial | syntax.json |
| cost | measured 2026 11 04 | cost |
| descriptant | present | epoch-memory |
| invariant | source derived | invariant |
| tests | executed bounded checks | tests |
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 |
|---|---|---|---|
school-student-vidtvorant-runner-behavior-complement | behavior_complement | partial source derived model | behavior-complement |
school-student-vidtvorant-runner-reverse-reconstruction | reverse_reconstruction | partial source derived model | reverse-reconstruction |
Model text
student_vidtvorant SCHOOL_DIR PROBE_ROOT OUT_DIR COURSE...: fits and measures the student's vidtvorant
(world/src/knowledge.rs.groupant/student-vidtvorant.groupant)
against student_probe traces.
forward
Load each course's traces from every repetition directory; refuse a course whose choices or understanding differ
between repetitions; take each question's median answer time. Read the lesson files' teaching lines. Parity folds
(fit on even lessons, test odd; the reverse): grid (anchor_min, own_min) ā {2,3}Ā², perceptron five epochs, gap gate
then no-gap gate from twenty score quantiles, best training agreement. Predict every question out of fold; fit the
time model per fold on the other fold's questions. Report agreement (choice, understood, correctness), per-course
scores, disagreement classes by the student's route (guess, support, support tie, rule or count) with examples, and
a course split (fit on courses 1, 3, 5ā¦, test the others). Then time predict per question with the first fold's
parameters.
invariant
semantic.json depends only on the traces' choices, scores and median times and on the lesson files: two runs are
byte-identical (checked four times, 2026-11-04). timing.json holds the nondeterministic own-speed numbers apart,
together with the deterministic time-model errors. No held-out set is read: the school's tests and exams are the
pupil's diagnostic score, and folds are by lesson parity or by course.
cost
ā 4 Ć (5 epochs + 40 gate replays) replays per fold, three folds; ā 170 s on one core for ten courses (7 412 questions).
behavior-complement
Does not run the student, the parser or any teacher; does not write outside OUT_DIR; does not register or claim a cap; does not select courses after seeing results (the course list is the command line).
reverse-reconstruction
A report is compatible with every set of traces giving the same per-question choices, scores and median times; the examples listed are a deterministic subset, not all disagreements.
tests
The included vidtvorant's six unit tests run with cargo test --release -p mova-approx --example student_vidtvorant; the repetition-consistency refusal and the semantic determinism are exercised by every run.
epoch-memory
2026-11-04: first version; results in $MOVA_DATA/runs/mova-speed-2026-11-04/vid-a..d.
This page in Markdown ā https://md.groupants.elamur.ai/free/en/mova/tools/approx/examples/student_vidtvorant.rs.groupant/