Elamur Groupantsquantitative-understanding
The groupant beside models/quantitative-understanding.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 | not materialized | |
| composition | partial | composition |
| cost | symbolic source derived | cost |
| descriptant | present | descriptant |
| invariant | partial | invariant |
| tests | see source local evidence | 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 |
|---|---|---|---|
behavior-complement | behavior_complement | partial source derived model | behavior-complement |
reverse-reconstruction | reverse_reconstruction | partial source derived model | reverse-reconstruction |
representation-transport | representation_transport | partial source derived model | representation-transport |
search-supervision | search_reachability | partial source derived model | search-supervision |
effect-boundary | effect_boundary | partial source derived model | effect-boundary |
cost-composition | resource_cost | partial source derived model | cost-composition |
Model text
descriptant
quantitative-understanding: wave28
qworld and deduct consume steps::Prob with aligned qs/descs. qworld rates change no cell immediately, but owner/thing context is updated before operation selection and can affect subsequent bare quantities. qread Set adds whereas qworld Set overwrites; qread's final event fold respects upto, while direct/relation/movement/initial-state decisions can inspect future events, so the complete function is not a prefix-state operator. steps oracle excludes repeated leaves; deduct gold can reuse a subtree but search excludes self-pairs. lex lookup requires caller normalization; understand does it. Missing-path test stops before Annotator load.
This source-bound partial group organizes existing Mova code. Native evidence and excluded effects are recorded in source sidecars and the root wave28 report. Registration is not a universal proof or whole-system runtime claim.
model
quantitative-understanding
forward
lex -> number/unit chunks -> placeholder owner projection -> ordered Interp frames/Pratt -> Expr -> evaluator. Independently steps produces aligned Prob quantities/descriptions; qworld folds exact rational cells and queries them; deduct converts reused quantity DAGs. qread uses separate f64 affine events with leaf provenance, constraint substitution, query and explicit abstain.
invariant
Source indices bind original quantity occurrences; arithmetic representations are distinct: exact Q cells, fixed Rat Expr and f64 candidate/affine computations. Query acceptance never establishes correctness of annotation. Model/teacher/corpus effects excluded from tests.
cost
Composition sums annotation, chunk/frame scanning, exact arithmetic, affine map/event scans, O(Nbeam20) stack expansions and O(stepsbeam(N+steps)^2) DAG candidates with feature/cloning/sorting overhead. Dense deduct training arrays2^22 entries each; no benchmark.
composition
qworld and deduct consume steps::Prob with aligned qs/descs. qworld rates change no cell immediately, but owner/thing context is updated before operation selection and can affect subsequent bare quantities. qread Set adds whereas qworld Set overwrites; qread's final event fold respects upto, while direct/relation/movement/initial-state decisions can inspect future events, so the complete function is not a prefix-state operator. steps oracle excludes repeated leaves; deduct gold can reuse a subtree but search excludes self-pairs. lex lookup requires caller normalization; understand does it. Missing-path test stops before Annotator load.
domain
Bounded synthetic valid structures and valid source indices, fixed environment; finite numeric interpretation where required. n<=64 for steps oracle and <=256 DAG pool indices. No external models, teacher, GPU, corpora, held-out or large training.
limitations
Dependency annotations and learned weights are not validated by helper fixtures. Public malformed structures may panic. No useful executable percent surrogate attached; future caps1/2/5/10 require metric.
checks
- 12 selected library wave28 tests plus1 exact missing-path integration test, twice debug and release
- Seven compiled assertion controls with exact byte restoration and focused positive reruns
- Seven syntax projections generated twice byte-identically
catalog
behavior-complement
Missing-path guard excludes loader effects; qworld rates exclude cell writes; valid DAG apply excludes invalid numeric interval; unknown-token Interp rejects before frame choice. These scoped exclusions do not compose into general English correctness.
reverse-reconstruction
Compatible source histories are constrained by quantity occurrence indices and selected frame/action semantics. Overwrites, placeholder loss, duplicate-valued matching and commutative keys make preimages set-valued; scalar output cannot reconstruct original text.
representation-transport
Exact rational cells cannot be equated with f64 search without conversion/error domain. Preserve source occurrence indices and distinct additive versus overwrite Set semantics.
search-supervision
steps oracle forbids repeated leaves; deduct gold reuses identical nodes but self-combination search candidates absent, so syntactically accepted gold need not be reachable.
effect-boundary
Only missing path metadata/tempdir effects exercised; Annotator and frozen suite remain excluded.
cost-composition
Feature extraction, candidate clone/tree construction and exact arithmetic costs survive composition; no measured speed or global percentage bound.
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/models/quantitative-understanding.groupant/