Elamur Groupantskg-textgen-order
The groupant beside kg/src/textgen.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 | partial | approximant |
| ast | partial | ../syntax.json |
| composition | partial | composition |
| cost | partial | cost |
| descriptant | present | descriptant |
| invariant | partial | 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 |
|---|---|---|---|
behavior_complement | behavior_complement | partial source derived model | behavior_complement |
reverse_reconstruction | reverse_reconstruction | partial source derived model | reverse_reconstruction |
class_alias | class_alias | partial source derived model | class_alias |
tie_policy | tie_policy | partial source derived model | tie_policy |
smoothing | smoothing | partial source derived model | smoothing |
position_tie | position_tie | partial source derived model | position_tie |
Model text
descriptant
kg-textgen-order
Bigram observations and subset maximum-score order.
Modeling epoch — 2026-10-31
Source read before tests and models; retained prospective notes accompany the parent. Current evidence uses bounded invented inputs only. This is partial modeling, not proof or corpus accuracy.
approximant
Forward relation
Observe sorted explicit heads, add start/end transitions. logp=ln((count+.1)/(from+3.2)). DP[mask,last] updates unused next and backpointers, final adds end transition. D[{i},i]=logp(<s>,c_i); for j∉S, D[S∪{j},j]=max_i∈S[D[S,i]+logp(c_i,c_j)]. Final score max_i D[All,i]+logp(c_i,</s>); backpointers recover maximizing permutation.
Executable approximant gap
Exact categorical structure, identifiers, parser decisions and effects have no useful percentage metric established here. No executable surrogate, automatic substitution or 1/2/5/10% guarantee is claimed.
invariant
Invariant and domain
For n1..12 and finite counts, recovered order is a head permutation maximizing this recurrence.
Cost and resources
O(n^2 2^n) time/O(n2^n) state; each score constructs string keys.
cost
Cost and resources
O(n^2 2^n) time/O(n2^n) state; each score constructs string keys.
composition
Composition
Bigram observations and subset maximum-score order.
Observe sorted explicit heads, add start/end transitions. logp=ln((count+.1)/(from+3.2)). DP[mask,last] updates unused next and backpointers, final adds end transition. D[{i},i]=logp(<s>,c_i); for j∉S, D[S∪{j},j]=max_i∈S[D[S,i]+logp(c_i,c_j)]. Final score max_i D[All,i]+logp(c_i,</s>); backpointers recover maximizing permutation.
The parent invokes scoped transformations through the actual functions named in their source ranges. Children partition useful observations, not a complete semantic proof; boundaries can omit helpers, tests and effects.
models
behavior_complement
No transition reuses a selected head. Private n0 is unsafe, and public generate guards it; no unbounded-n contract.
reverse_reconstruction
Order corresponds to all count tables with same chosen optimum/ties; observations cannot reconstruct sentence histories uniquely.
class_alias
Prep and quantity details collapse through class before observation; objects retain case.
tie_policy
Strict > retains first predecessor, last equal final max endpoint selected.
smoothing
Unseen transition count0 remains finite log probability with alpha.1 and fixed32 classes.
position_tie
Heads sorted by tuple(word,class); tied coordinates use lexical class, not original member order.
tests
Evidence
Parent source checks (a fragment inherits only the behavior they actually exercise):
- textgen::tests::subject_verb_object
- textgen::wave29::variant_priority_and_head_bounds
Selected kg tests: 15 named outcomes, twice debug/twice release, equal sorted outcomes. Standalone actual global/build.rs: 3 named outcomes per run, twice debug/twice optimized; build_child is normally no-op, parent fixture invokes it against isolated synthetic seeds. Eight compiled source mutations (one per source) each compiled successfully and exited101 on the focused assertion; exact source restoration followed by positive rerun.
This fragment's untested branches remain source-derived. No live graph, DB, corpus, teacher, GPU or held-out evaluation was used. Command specs and numeric exits are retained in the external wave29 graphlang receipt. Strict kg Clippy reports existing source lints; see receipt.
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/kg/src/textgen.rs.groupant/order.groupant/