145d797196
Measurement #1 of the field-reasoner falsifiable experiment: does the CL(4,1) field, given an honest metric encoding, read forward-substitutable quantitative-relational problems from TEXT with wrong==0? It does — 14/15 correct, 0 wrong, 1 refused (precision ceiling), scored against an independent arithmetic oracle. - generate/relational_field_reader.py: reads problem text into conformal points on the e1 number line; additive/part-whole relations are conformal TRANSLATOR versors (versor_apply(T_delta, embed[x]) == embed[x+delta], exact); the answer reads back by projective dehomogenization. Refusal-first: fences multiplicative/ratio (the sign/orientation-blind cases), the precision ceiling, non-forward-substitutable references, negatives. A per-step exactness self-check turns any f64 translator drift into a refusal (precision_drift) — it NEVER commits a wrong integer. Its parser is an independent reimplementation importing no generate.derivation/math_*. - evals/relational_metric/: independent arithmetic oracle (computes gold from the STRUCTURE, shares no code with the reader), 15-case fixture, and a runner that enforces gold integrity + wrong==0. - INV-25: relational_metric registered in INDEPENDENT_GOLD_LANES (oracle proven code-disjoint from the field reader and the algebra engine). The independently golded panel is now three domains: deductive, dimensional, relational-metric. Green: smoke 87, 53 architectural invariants, 16 new tests; deductive + dimensional lanes unperturbed (wrong=0).
35 lines
1.5 KiB
Python
35 lines
1.5 KiB
Python
"""Relational-metric gold lane — the field reader is wrong==0 vs an independent gold.
|
|
|
|
This is measurement #1 of the field-reasoner wedge falsifiable experiment: does the
|
|
geometric field reader, reading problem TEXT, commit answers that match an
|
|
independent arithmetic oracle (computed from the STRUCTURE) with zero wrong? The
|
|
oracle shares no code with the reader (enforced structurally by INV-25's
|
|
INDEPENDENT_GOLD_LANES registration).
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
from evals.relational_metric.runner import run
|
|
|
|
|
|
def test_lane_is_wrong_zero_with_independent_gold():
|
|
report = run()
|
|
# The committed gold is reproducible by the independent oracle (never field-derived).
|
|
assert report["gold_integrity_failures"] == [], report["gold_integrity_failures"]
|
|
# wrong==0 is the prime directive on this lane.
|
|
assert report["wrong"] == 0, report["wrong_detail"]
|
|
# Buckets account for every case.
|
|
assert report["total"] == report["correct"] + report["wrong"] + report["refused"]
|
|
|
|
|
|
def test_field_actually_commits_not_a_refusal_floor():
|
|
"""A refusal floor (refuse everything) would be wrong==0 too — and worthless.
|
|
The capability claim is COVERAGE with wrong==0: the field must commit real cases."""
|
|
report = run()
|
|
assert report["correct"] >= 10
|
|
|
|
|
|
def test_over_ceiling_is_refused_not_wrong():
|
|
"""The precision ceiling is honest coverage (a refusal), never a wrong commit."""
|
|
report = run()
|
|
assert any("over_ceiling" in r for r in report["refused_detail"])
|