Elamur Groupantsanswer-span-normalization
The groupant beside world/src/span.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 source model | forward |
| ast | partial | ../syntax.json |
| cost | partial source model | cost |
| descriptant | partial source model | epoch-memory |
| invariant | partial source model | invariant |
| tests | partial source model | checks |
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 |
|---|---|---|---|
answer-span-normalization-behavior-complement | behavior_complement | partial source model | behavior-complement |
answer-span-normalization-reverse-reconstruction | reverse_reconstruction | partial source model | reverse-reconstruction |
answer-span-normalization-echo-threshold | threshold_relation | partial source model | echo-threshold |
answer-span-normalization-pronoun-vacuity | constraint_relation | partial source model | pronoun-vacuity |
answer-span-normalization-finite-list-fibers | reverse_reconstruction | bounded native correspondence | finite-list-fibers |
answer-span-normalization-marker-boundary-intervention | boundary_relation | bounded native correspondence | marker-boundary-intervention |
Model text
model
answer-span-normalization
Subject: world/src/span.rs, whole-file SHA-256 in manifest, baseline 4279da541. Domain: finite acyclic valid trees with consistent child lists, valid indices and terminating dependencies. Observation: a single call return, not semantic truth or reachability. Hold Tree, ans::lemmas and types::FEELING dependencies fixed; parser/model loading and external caller wiring are gaps. All entries are source-derived partial models. Exact-target definitions are not executable inverses; finite tests cannot establish universal correspondence.
forward
Sort/dedup, asymmetric edge trim, then content/echo gate; is_pron is a separate predicate.
invariant
Interior EDGE tokens remain; valid indices required.
cost
O(m log m+m^2) time, O(m) auxiliary content.
epoch-memory
2026-10-29: initial source-bound epoch. Preserve Num bypass, global Feel selection, early-return cleanup suppression, and verb_kids comment mismatch: code expands Xcomp only and checks no shared subject. No source changes or task-accuracy claim. Partial syntax has no semantic analysis or macro expansion.
checks
Structural inventory checks hashes, references, types and node bounds only. Proposed oracle: enumerate small annotated acyclic trees and compare literal expected returns, including None and labels; mutate priority or echo inequality to obtain a red control. Any executed bounded evidence is recorded separately in root evidence.md; catalog examples remain proposed until recorded there.
composition
Leaf scope shares immutable tree/question/helper contracts. Overlap with parent is not extra runtime work. Dependencies outside range remain opaque.
behavior-complement
U=(Tree,valid indices,qlem) x optional indices; R=clean. Inner U minus R excludes success unsorted/duplicate, content-free, majority-echo, leading EDGE/Cc/Punct/Mark or trailing EDGE/Cc/Punct. Non-EDGE trailing Mark is not excluded.
Assumptions/evidence/status: common domain above; source-derived partial relation at bound revision. Approximation and held-fixed inputs are stated in the relation. Oracle: finite literal fixtures exercising both membership and nonmembership against the actual function, preserving ambiguous preimages. No linguistic correctness or universal proof inferred.
reverse-reconstruction
For J retain all index lists whose sorting/dedup and edge trim gives J and echo gate passes. Duplicates/permutations/trimmed edges arbitrary within that constraint. None fibers include no-content and majority-echo candidates.
Assumptions/evidence/status: common domain above; source-derived partial relation at bound revision. Approximation and held-fixed inputs are stated in the relation. Oracle: finite literal fixtures exercising both membership and nonmembership against the actual function, preserving ambiguous preimages. No linguistic correctness or universal proof inferred.
echo-threshold
For c>0 content and e echoed content, success iff 2e<=c; equality c=2,e=1 passes. c=0 separately fails. Adding qlem membership can cross threshold; echo==c test is redundant when c>0.
Assumptions/evidence/status: common domain above; source-derived partial relation at bound revision. Approximation and held-fixed inputs are stated in the relation. Oracle: finite literal fixtures exercising both membership and nonmembership against the actual function, preserving ambiguous preimages. No linguistic correctness or universal proof inferred.
pronoun-vacuity
is_pron is universal PRON/DET/PUNCT membership, true for empty indices. It is separate from clean: mixed PRON+NOUN is not pronoun-only but may fail echo. Selected subject/object/speech gates use it before cleanup.
Assumptions/evidence/status: common domain above; source-derived partial relation at bound revision. Approximation and held-fixed inputs are stated in the relation. Oracle: finite literal fixtures exercising both membership and nonmembership against the actual function, preserving ambiguous preimages. No linguistic correctness or universal proof inferred.
wave6
Native selection observations, 2026-10-30
Revision: exact source SHA in this manifest. Observations are single terminating calls. Tree fixtures are directly annotated, acyclic and index-valid; dependencies are compiled from this checkout. No parser or teacher is loaded. Metric is exact equality of optional indices/text and rule labels; zero tolerance. Exhaustive claims apply only to explicitly bounded domains, with no semantic accuracy or general Rust proof.
finite-list-fibers
Fix tree [and/Cc, ruby/NOUN, jade/NOUN, ./Punct]. Domain D is every index list of length 0..4 over {0,1,2,3}, crossed with four question-membership masks over ruby/jade: 1364 calls. Let S be ascending unique input indices intersected with {1,2}. A(S,q) is Some(S) for S=[1] with ruby absent, S=[2] with jade absent, or S=[1,2] unless both present; otherwise None. For every output y, B_q(y)={L in D_lists:A(L,q)=y}; complete enumeration compares every member of each production fiber with this independently specified truth table, including None. Ambiguous duplicate/permuted/edge antecedents remain. Test cleanup_complete_bounded_fibers supplies exact bounded correspondence, not an unbounded reconstruction algorithm. Cost: 341 list evaluations per mask, maximum four indices each.
marker-boundary-intervention
Fix q empty and tree [to/Mark, ruby/NOUN, and/Cc, jade/NOUN, to/Mark]. For all-index candidate plus duplicate ruby, normalization returns [1,2,3,4]: leading non-EDGE Mark disappears, trailing non-EDGE Mark and internal Cc survive. Subdomains [0,1] and [1,4] yield [1] and [1,4]. Model forbids symmetric Mark stripping on these calls (inner complement) and retains the original indices as compatible antecedents, without claiming these three lists exhaust the full fiber. Test cleanup_marker_asymmetry_and_internal_edges observes indices, so repeated spelling cannot hide boundary errors.
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/world/src/span.rs.groupant/normalization.groupant/