Elamur Groupantsmath-curriculum
The groupant beside math/src/curric.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 |
Parts
Model text
descriptant
Source-bound description
sample stable-sorts splitmix(SEED xor fnv(salt:id)) then takes k. prep assigns floor(n/2) dev before deduplicating ids (reported counts may exceed rows). grade recognized option overrides answer; else letter or unique numeric match. Int rational equality exact; otherwise abs error<=1%*max(abs(gold),1e-9) OR<1e-9.
Epoch 2026-10-31: prospective model before bounded synthetic tests; production behavior unchanged. No corpus prep/load or held grading; arbitrary malformed numeric normalization not certified.
approximant
Forward relation
sample stable-sorts splitmix(SEED xor fnv(salt:id)) then takes k. prep assigns floor(n/2) dev before deduplicating ids (reported counts may exceed rows). grade recognized option overrides answer; else letter or unique numeric match. Int rational equality exact; otherwise abs error<=1%*max(abs(gold),1e-9) OR<1e-9.
Domain: Synthetic samples/Problems only. MMLU indexing assumes<=5 choices and valid index; malformed input can panic. No external prep.
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
Recognized wrong option excludes correct grade even with right numeric answer; unknown option permits fallback. k=0 returns empty. NaN does not pass tolerance.
Ordering: Recognized option→short letter→numeric unique match; norm_num tries rational then first number then expression.
Cost: O(n log n)+clone/hash bytes sampling, complete JSONL pools for prep. Grading parsing/evaluation and big arithmetic delegated.
composition
Composition and epoch memory
Problem kind/source→solve Expect; option_value and norm_num differ, so not identical normalization.
No corpus prep/load or held grading; arbitrary malformed numeric normalization not certified.
Partial syntax and finite tests do not establish all caller assumptions.
models
behavior_complement
Recognized wrong option excludes correct grade even with right numeric answer; unknown option permits fallback. k=0 returns empty. NaN does not pass tolerance.
reverse_reconstruction
Sample admits all pools with same keyed prefix and unseen tail; collisions retain input order. Grade true admits numeric tolerance neighborhood or alternate spellings, not unique answer.
failure_precedence
Recognized option→short letter→numeric unique match; norm_num tries rational then first number then expression.
representation_transport
Problem kind/source→solve Expect; option_value and norm_num differ, so not identical normalization.
resource_boundary
O(n log n)+clone/hash bytes sampling, complete JSONL pools for prep. Grading parsing/evaluation and big arithmetic delegated.
domain_boundary
Synthetic samples/Problems only. MMLU indexing assumes<=5 choices and valid index; malformed input can panic. No external prep.
tests
Evidence, 2026-10-31
Selected actual-source checks: cargo test -p math --lib curric::. Exact commands, numeric statuses, normalized outcomes, compiled assertion controls and restoration receipts are in the external wave28 mathcontrol run.
No corpus prep/load or held grading; arbitrary malformed numeric normalization not certified.
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/curric.rs.groupant/