ad7993e861
Completes the Phase 5 curriculum-era lane checklist alongside 5.1.
English-substrate domain lanes (5.4–5.7) — extend the proven
english_fluency_ood pattern with new vocabulary domains. Same
13-construction realizer, same grammatical_coverage rubric, new
triples. All four lanes land at 100% on both splits:
5.4 elementary_mathematics_ood 117/117 public + 39/39 holdouts
domains: arithmetic, set, geometry | holdout: probability
5.5 foundational_physics_ood 117/117 + 39/39
domains: mechanics, electricity, thermodynamics | holdout: optics
5.6 foundational_biology_ood 117/117 + 39/39
domains: cell, organism, ecosystem | holdout: genetics
5.7 classical_literature_ood 117/117 + 39/39
domains: epic, tragedy, lyric | holdout: comedy
New-language lanes (5.2 Hebrew, 5.3 Koine Greek) — scoped honestly to
v1 = C01 only, script + length rubric. The realizer's
tense/aspect/quantifier/negation logic in generate/templates.py is
English-only; C02-C13 in HE/GRC requires Hebrew/Greek morphology +
rhetorical templates, named explicitly in each lane's gaps.md as the
v2 unblock path. v1 measures what infrastructure exists:
5.2 hebrew_fluency 3/3 (predicate-subject-object assembly,
Hebrew script gate)
5.3 koine_greek_fluency 3/3 (subject-object-predicate assembly,
Greek script gate)
Lane scaffolds follow the established pattern: contract.md, runner.py,
__init__.py, gaps.md, public/v1/cases.jsonl, dev/cases.jsonl,
holdouts/v1/cases.jsonl (5.4–5.7 only; HE/GRC holdouts deferred to v2
when vocabulary expands).
Generators + scorers:
scripts/generate_phase5_domain_lanes.py — 5.4–5.7 case emit
scripts/scaffold_phase5_domain_lanes.py — 5.4–5.7 contracts/runners
scripts/generate_phase5_language_lanes.py — 5.2/5.3 case emit
scripts/score_phase5_holdouts.py — parallel holdouts scoring
via multiprocessing.Pool
(mirrors the parallel-eval
pattern from evals/parallel.py)
Lanes are wired into core eval --list automatically through the
framework's lane discovery; parallel sweeps via bash background jobs
(one process per lane).
Regression clean: smoke 54, runtime 19, teaching 17, packs 6,
cognition 57, algebra 132. Cognition eval 100% across all metrics.
2.9 KiB
Koine Greek fluency — gaps
v1 (current)
- Construction coverage: C01 only (simple declarative).
- Word order: subject-object-predicate, via
generate.articulation._assembleforlanguage == "grc". - Grounding: through
ChatRuntime.chat()withframe_pack="grc". - Rubric: script + length only. Lexeme-level slot matching is not gated at v1 for the same runtime-folding reason as Hebrew (see v2 §6 below).
v2 unblock path — Koine Greek construction coverage
To extend to C02–C13 (negation, tense, aspect, quantification, relative clause, etc.), the realizer needs Greek analogues of the English-only modules:
-
Greek morphology — analogue of
generate/morphology.py. At minimum: verb conjugation across the principal parts (present, future, aorist, perfect, perfect middle, aorist passive), participles (active/middle/passive across tenses), noun declension (1st/2nd/3rd) with case-marked subject/object slots instead of word-order-marked. -
Greek predicate display map — analogue of
_PREDICATE_DISPLAYingenerate/templates.pymapping seed-pack predicates to surface forms with appropriate aspect marking. -
Greek rhetorical templates — analogue of
_MOVE_TEMPLATES. Particle-driven discourse markers (μέν / δέ, γάρ, οὖν) instead of English function words ("furthermore", "in contrast"). -
Greek negation — οὐ / μή placement before the verb, with the οὐ/μή split governed by mood (indicative vs. non-indicative).
-
Greek quantifiers — πᾶς (all), τις (some/indefinite), with declension matching the subject's case/number/gender.
-
Case-marked agreement — Greek's free word order is governed by morphological case; the realizer needs to mark subject in nominative and object in accusative regardless of surface order. The lexicon entries currently lack case-paradigm tags.
-
Lexeme-level slot grounding — same limitation as Hebrew: the GRC runtime pipeline currently produces single-lexeme articulations for multi-token Greek input. v2 needs the grounding + planning layers to preserve distinct subject/predicate/object slots so the rubric can check lexeme presence per slot.
Out of scope for this lane
- Polytonic accent generation. Lexicon entries carry accents as fixed strings; the realizer does not currently compute accent shift on enclisis.
- Attic vs. Koine register distinction. Seed pack is Koine (logos-tier vocabulary); Attic expansion is a separate pack.
- Reconstructive pronunciation (Erasmian vs. modern vs. reconstructed Koine). Lane scores the script; pronunciation is out of band.
Related
- ADR-0020 (Rust parity sequencing) — language-track work is independent of the Rust parity track.
- Phase 5.2 (Hebrew) — same v1 scope rationale; same v2 unblock pattern, swapped for Greek-specific morphology.