Elamur Groupantsmath-solver-cli-training-routes
The groupant beside math/src/bin/mathsolve.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 |
| descriptant | present | descriptant |
| invariant | partial | invariant |
| tests | partial bounded evidence | 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 |
failure_precedence | failure_precedence | partial source derived model | failure_precedence |
representation_transport | representation_transport | partial source derived model | representation_transport |
resource_boundary | resource_boundary | partial source derived model | resource_boundary |
domain_boundary | domain_boundary | partial source derived model | domain_boundary |
Model text
descriptant
Source-bound description
steps/arith load model/datasets, construct gold trees, train across epochs/folds, optionally rerank/ensemble and assess. steps optional hints calls teacher after disagreement selection. arith chooses gold trees using current model.
Epoch 2026-10-31: prospective model before bounded synthetic tests; production behavior unchanged. Source scope bytes[7712,66119). Large effect branches source-modeled only; controls command avoided because potentially expensive Fermat expression. Not every branch in this fragment executed.
approximant
Forward relation
steps/arith load model/datasets, construct gold trees, train across epochs/folds, optionally rerank/ensemble and assess. steps optional hints calls teacher after disagreement selection. arith chooses gold trees using current model.
Domain: Entire branch unexecuted; external corpora/model and explicit authorization required for actual training/held workloads.
Executable Rust approximant unsupported: categorical/effect observations have no useful percentage metric yet. Future caps1/2/5/10% only; no automatic substitution.
invariant
Invariant and cost
Pure flag tests exclude no corpus/model side effects of these branches. Exact gold filtering may exclude training cases but not ensure test correctness.
Ordering: Preparation/gold filtering before training, averaging before assessment; optional hints after collecting cases.
Cost: Datasetepochssearch complexity; extra folds/bagging/STaR multiply work, memory holds prepared examples/candidates. No bounded-run claim.
composition
Composition and epoch memory
Delegates steps/deduct/arith/qworld/qread/global; implementation and data policy not proven by CLI source model.
Source scope bytes[7712,66119). Large effect branches source-modeled only; controls command avoided because potentially expensive Fermat expression. Not every branch in this fragment executed.
Partial syntax and finite tests do not establish all caller assumptions.
models
behavior_complement
Pure flag tests exclude no corpus/model side effects of these branches. Exact gold filtering may exclude training cases but not ensure test correctness.
reverse_reconstruction
Printed accuracy admits many prediction vectors; learned weights and gold supervision cannot be reconstructed from summary percentages.
failure_precedence
Preparation/gold filtering before training, averaging before assessment; optional hints after collecting cases.
representation_transport
Delegates steps/deduct/arith/qworld/qread/global; implementation and data policy not proven by CLI source model.
resource_boundary
Datasetepochssearch complexity; extra folds/bagging/STaR multiply work, memory holds prepared examples/candidates. No bounded-run claim.
domain_boundary
Entire branch unexecuted; external corpora/model and explicit authorization required for actual training/held workloads.
tests
Evidence, 2026-10-31
Selected actual-source checks: cargo test -p math --bin mathsolve wave28; isolated calc smoke. Exact commands, numeric statuses, normalized outcomes, compiled assertion controls and restoration receipts are in the external wave28 mathcontrol run.
Source scope bytes[7712,66119). Large effect branches source-modeled only; controls command avoided because potentially expensive Fermat expression. Not every branch in this fragment executed.
No live teacher/API, trained model, corpus/held-out, GPU, DB or expensive benchmark run. Parent syntax generated twice identically. Scope remains partial.
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/math/src/bin/mathsolve.rs.groupant/training-routes.groupant/