The confuser probe's two pseudo-accumulation misfires (0005 ->796, 0007 ->996)
both traced to the same extraction blind spot: a number bonded to its unit by a
hyphen (`25-foot sections`, `20-inch pieces`) was invisible to the base
`number + space + word` pattern, so the self-verification completeness clause
never saw the divisor and the bare `buys ... gives` accumulation read as
"complete". This is the highest-leverage lever the 2026-05-29 session named
("the gate must see the fractions/25-foot it currently misses").
EX-6 adds a tight, ADR-0165-safe lexeme pass: a digit run, a single hyphen, an
alphabetic unit word. The alphabetic-only unit group keeps numeric ranges (`3-5`)
out; taking only the first hyphen segment keeps the postmodifier tail
(`25-year-old`) from inflating the unit — so it stays clear of the deferred EX-3
multi-word-unit traps. Over-extraction here is strictly refuse-preferring: making
the divisor visible drives 0005/0007 to refuse via the polarity-None
`cuts`/`splits` clause, never to a wrong answer.
Evidence (deterministic, the microscope):
- confuser probe: wrong 7 -> 5; pseudo-accumulation 0 wrong / 4 refused;
genuine positives still 7 solved; pair-tells unchanged (4).
- train_sample (capability): 3/47/0 byte-identical.
- practice accumulation: 3/47/0 (wrong=0) byte-identical.
- smoke 67 passed; lane-SHA freeze 8/8 (serving frozen).
Tests:
- TestEX6HyphenatedUnitNumbers pins the new lexeme (value+unit, decimal, no
double-count, word-compound unaffected, numeric range not read as a unit).
- TestProbeBaseline tightened wrong 7->5; new test_pseudo_accumulation_does_not_misfire
is the failing-under-violation obligation (fails loudly if the pass regresses).
- TestSlashFractionLeakHazard pins the deferred `1/4`->`4` denominator leak: not
fixed here because suppressing the leaked operand *removes* a quantity and can
unblock the completeness clause (not unambiguously refuse-preferring), so it
needs its own train_sample + probe validation.
248 lines
12 KiB
Python
248 lines
12 KiB
Python
"""ADR-0175 Phase 3b / ADR-0179 — lexeme-level quantity extraction.
|
|
|
|
Pulls ``(value, unit, source_token)`` triples from a problem using conservative
|
|
orthographic patterns. Per ADR-0165 these are *lexeme* patterns ("what this
|
|
piece looks like: a number, a unit word") — never grammar templates ("how words
|
|
combine to mean X"). The *combining* is the search's job (search.py / compose.py)
|
|
gated by self-verification, which is refuse-preferring; over-extraction here can
|
|
only cost *refusals*, never a wrong answer.
|
|
|
|
ADR-0179 enrichments integrated here (sealed lane only — ``chat/`` does not import
|
|
this module, so none of this can move the serving ``3/47/0``):
|
|
|
|
* **EX-1 — word-numbers.** ``"three apples"`` → ``3.0``, including tens-one
|
|
hyphen compounds (``"twenty-four"`` → ``24.0``). Reuses the canonical
|
|
``WORD_NUMBERS`` table from :mod:`generate.math_roundtrip` (single number
|
|
vocabulary). Factor-bearing forms (``half``/``third``/``quarter``) are excluded
|
|
— they read as divisors, not counts.
|
|
* **EX-4 — list-unit inheritance.** In a bare numeric list with the unit stated
|
|
once at the end (``"20, 36, 40 and 50 push-ups"``) the trailing unit attaches
|
|
to every number in the list. Still orthographic: a run of number tokens joined
|
|
by comma/``and`` delimiters, then one unit token. Whether the resulting
|
|
quantities may compose is the gate's decision, not the extractor's.
|
|
* **EX-5 — sentence-final numbers.** A number with no following unit word (end of
|
|
sentence/text or before terminal punctuation) extracts with an empty unit so it
|
|
stays available to the completeness check without inventing a unit lexeme.
|
|
* **Unit hygiene (function-word filter).** When the token after a number is a
|
|
function word (``$0.75 each`` → ``each``, ``$40 to go`` → ``to``, ``3/4 of`` →
|
|
``of``), the single-word unit pattern would tag it as the unit — a spurious unit
|
|
that corrupts same-unit detection (GB-2/GB-3) and CP-1's ``unit_shape``. Such
|
|
units are blanked (empty, like a sentence-final number): the value stays
|
|
grounded, the unit is honestly unknown. Closed lexeme set (``_NON_UNIT_WORDS``),
|
|
not a grammar template (ADR-0165).
|
|
* **EX-6 — hyphen-bonded number-units** (ADR-0163-F2). A number bonded to its unit
|
|
by a hyphen (``25-foot``, ``20-inch``, ``2.5-mile``) was invisible to the base
|
|
``number + space + word`` pattern, so the completeness check never saw the
|
|
divisor — the blind spot behind the pseudo-accumulation confusers (0005/0007).
|
|
Tight lexeme (digit run, hyphen, alphabetic unit word); the alphabetic-only unit
|
|
group keeps numeric ranges (``3-5``) out and only the first hyphen segment is
|
|
taken, so it stays clear of the deferred EX-3 multi-word-unit traps below.
|
|
|
|
EX-3 (multi-word units) is deliberately **not** integrated. Two distinct traps
|
|
defeat the tightest lookahead-anchored rule the brief admits:
|
|
|
|
1. **Connective-crossing** (in
|
|
``docs/handoff/AUDIT-ADR-0179-EX-RECONCILE.md``). The greedy lowercase unit
|
|
span regresses GB-2's same-unit detection (``"6 apples and 4 apples"`` → unit
|
|
``"apples and"``) and does not cleanly recover real multi-word units from
|
|
0024-class text (``"20 jumping jacks on Monday"`` → ``"jumping jacks on"``).
|
|
2. **Postmodifier-adjective tails** (discovered during the Track C redo of
|
|
``docs/handoff/PARALLEL-WORK-PLAN-2026-05-29.md``). Even a *tight*
|
|
``digit + lc word + lc word + (?=clause-terminator)`` rule fires on
|
|
``"25 years old?"`` and produces unit ``"years old"`` instead of
|
|
``"years"`` — regressing
|
|
``tests/test_adr_0176_ms1_question_target.py::TestQuestionQuantities::test_extracts_quantity_stated_in_question``.
|
|
The pattern is endemic: GSM8K cases 0006/0033 and several MS2 chain tests use
|
|
``"X years old"``. Closing it would need a second closed lexeme set (a stop
|
|
list of measurement postmodifiers — ``old``, ``tall``, ``long``, ``wide``,
|
|
``deep``, ``high``, ``away``, ``apart``, ``ago``, …) which the audit did not
|
|
anticipate and which the Track C brief judged too open-ended to enumerate
|
|
responsibly. ``TestEX3StillDeferred`` in
|
|
``tests/test_adr_0179_extract.py`` pins this second trap so no future redo
|
|
silently re-introduces it.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import re
|
|
from typing import Final
|
|
|
|
from generate.derivation.model import Quantity
|
|
from generate.math_roundtrip import WORD_NUMBERS
|
|
|
|
# Number (int or decimal) immediately followed by a single unit word. Lexeme-level.
|
|
_QTY_RE: Final[re.Pattern[str]] = re.compile(
|
|
r"(?<![\w.])(\d+(?:\.\d+)?)\s+([a-zA-Z]+)"
|
|
)
|
|
|
|
# EX-4: a same-unit numeric list with the unit stated once at the end, e.g.
|
|
# ``20, 36, 40 and 50 push-ups``. A run of number tokens separated by comma/and
|
|
# delimiters, followed by one (optionally hyphenated) unit token.
|
|
_LIST_WITH_TRAILING_UNIT_RE: Final[re.Pattern[str]] = re.compile(
|
|
r"(?<![\w.])"
|
|
r"((?:\d+(?:\.\d+)?\s*(?:,\s*|\s+and\s+)){1,}\d+(?:\.\d+)?)"
|
|
r"\s+([a-zA-Z]+(?:-[a-zA-Z]+)*)"
|
|
)
|
|
_NUMBER_RE: Final[re.Pattern[str]] = re.compile(r"\d+(?:\.\d+)?")
|
|
|
|
# EX-1: word-number (optionally a tens-one hyphen compound) followed by a unit
|
|
# word. Built from the round-trip table; factor-bearing forms excluded.
|
|
_WORD_NUMBER_ALT: Final[str] = "|".join(
|
|
re.escape(word)
|
|
for word in sorted(WORD_NUMBERS, key=len, reverse=True)
|
|
if word not in {"half", "third", "quarter"}
|
|
)
|
|
_WORD_QTY_RE: Final[re.Pattern[str]] = re.compile(
|
|
rf"(?i)\b({_WORD_NUMBER_ALT})(?:-({_WORD_NUMBER_ALT}))?\s+([a-zA-Z]+)\b"
|
|
)
|
|
|
|
# EX-5: a number with no following unit, at end of sentence/text.
|
|
_FINAL_NUMBER_RE: Final[re.Pattern[str]] = re.compile(
|
|
r"(?<![\w.])(\d+(?:\.\d+)?)(?=\s*(?:[.?!]|$))"
|
|
)
|
|
|
|
# EX-6: a number bonded to its unit by a hyphen (``25-foot``, ``20-inch``,
|
|
# ``2.5-mile``). The base ``_QTY_RE`` requires whitespace before the unit word,
|
|
# so a hyphen-bonded unit was invisible — the blind spot behind the
|
|
# pseudo-accumulation confusers (``25-foot`` / ``20-inch`` divisors the
|
|
# completeness check never saw). Tight lexeme: a digit run, a single hyphen, an
|
|
# alphabetic unit word. The trailing ``[a-zA-Z]+`` (not ``\d``) keeps numeric
|
|
# ranges (``3-5``) out, and taking only the first hyphen segment keeps the
|
|
# postmodifier tail (``25-year-old`` -> unit ``year``) from inflating the unit —
|
|
# so this stays clear of the deferred EX-3 multi-word-unit traps.
|
|
_HYPHEN_QTY_RE: Final[re.Pattern[str]] = re.compile(
|
|
r"(?<![\w.])(\d+(?:\.\d+)?)-([a-zA-Z]+)"
|
|
)
|
|
|
|
|
|
# Function words that are never units. When the token immediately after a number
|
|
# is one of these (``$0.75 each``, ``$40 to go``, ``3/4 of``), the single-word unit
|
|
# pattern would otherwise tag the function word as the unit — a spurious unit that
|
|
# corrupts same-unit detection (GB-2/GB-3) and CP-1's unit_shape. Emitting an empty
|
|
# unit instead (like a sentence-final number) is honest: the value is grounded, the
|
|
# unit is simply unknown. Closed lexeme set (cf. ``WORD_NUMBERS``); ADR-0165-safe —
|
|
# it names tokens that are not unit nouns, it does not parse sentence structure.
|
|
_NON_UNIT_WORDS: Final[frozenset[str]] = frozenset(
|
|
{
|
|
"a", "an", "the", "of", "to", "for", "in", "on", "at", "as", "than",
|
|
"per", "each", "every", "and", "or", "with", "by", "from", "more",
|
|
"less", "about", "that",
|
|
}
|
|
)
|
|
|
|
|
|
def _clean_unit(unit: str) -> str:
|
|
"""Lowercase a unit token; blank it if it is a non-unit function word."""
|
|
lowered = unit.lower()
|
|
return "" if lowered in _NON_UNIT_WORDS else lowered
|
|
|
|
|
|
def _quantity(value_token: str, unit: str) -> Quantity | None:
|
|
"""Build a quantity from an already-matched numeric token."""
|
|
try:
|
|
value = float(value_token)
|
|
except ValueError: # pragma: no cover - regex guarantees numeric
|
|
return None
|
|
return Quantity(value=value, unit=_clean_unit(unit), source_token=value_token)
|
|
|
|
|
|
def _resolve_word_number(first: str, second: str | None) -> float | None:
|
|
"""Resolve a conservative word-number token from ``WORD_NUMBERS``.
|
|
|
|
A bare word resolves directly. A hyphen compound resolves only as a tens-one
|
|
form (``twenty-four``, ``ninety-nine``); anything else returns ``None`` so the
|
|
extractor stays conservative rather than guessing a composition rule.
|
|
"""
|
|
first_value = WORD_NUMBERS.get(first.lower())
|
|
if first_value is None:
|
|
return None
|
|
if second is None:
|
|
return float(first_value)
|
|
second_value = WORD_NUMBERS.get(second.lower())
|
|
if second_value is None:
|
|
return None
|
|
if first_value < 20 or first_value >= 100 or not 0 < second_value < 10:
|
|
return None
|
|
return float(first_value + second_value)
|
|
|
|
|
|
def _claimed(pos: int, spans: list[tuple[int, int]]) -> bool:
|
|
"""Whether a numeric token at ``pos`` was already claimed by an earlier pass."""
|
|
return any(start <= pos < end for start, end in spans)
|
|
|
|
|
|
def extract_quantities(problem_text: str) -> tuple[Quantity, ...]:
|
|
"""Extract ``(value, unit, source_token)`` quantities in left-to-right order.
|
|
|
|
Deterministic. ``source_token`` is the surface number string (used by the
|
|
self-verification gate to prove the value is grounded in the text). Units are
|
|
lowercased; the value's surface token is preserved verbatim.
|
|
|
|
Passes run most-specific first and claim the digit spans they consume so later
|
|
passes never double-count a number:
|
|
|
|
1. EX-4 same-unit list (claims every number in the list);
|
|
2. EX-6 hyphen-bonded number-unit (``25-foot``; claims the digit span);
|
|
3. digit + single unit word (skips numbers a list/hyphen pass already claimed);
|
|
4. EX-1 word-number + unit word (alphabetic, disjoint from digit spans);
|
|
5. EX-5 sentence-final bare number (skips any already-claimed digit).
|
|
"""
|
|
found: list[tuple[int, Quantity]] = []
|
|
claimed: list[tuple[int, int]] = []
|
|
|
|
# 1. EX-4 — list with one trailing unit; the unit inherits to every number.
|
|
for match in _LIST_WITH_TRAILING_UNIT_RE.finditer(problem_text):
|
|
unit = match.group(2).lower()
|
|
for num in _NUMBER_RE.finditer(match.group(1)):
|
|
pos = match.start(1) + num.start()
|
|
quantity = _quantity(num.group(0), unit)
|
|
if quantity is not None:
|
|
found.append((pos, quantity))
|
|
claimed.append((pos, pos + len(num.group(0))))
|
|
|
|
# 1b. EX-6 — hyphen-bonded number-unit (``25-foot``). Claims the digit span so
|
|
# the bare/final passes never re-surface the number with a blank unit.
|
|
for match in _HYPHEN_QTY_RE.finditer(problem_text):
|
|
if _claimed(match.start(1), claimed):
|
|
continue
|
|
quantity = _quantity(match.group(1), match.group(2))
|
|
if quantity is not None:
|
|
found.append((match.start(1), quantity))
|
|
claimed.append(match.span(1))
|
|
|
|
# 2. digit + single unit word — the original base pattern.
|
|
for match in _QTY_RE.finditer(problem_text):
|
|
if _claimed(match.start(1), claimed):
|
|
continue
|
|
quantity = _quantity(match.group(1), match.group(2))
|
|
if quantity is not None:
|
|
found.append((match.start(1), quantity))
|
|
claimed.append(match.span(1))
|
|
|
|
# 3. EX-1 — word-numbers (and tens-one hyphen compounds) with a unit word.
|
|
for match in _WORD_QTY_RE.finditer(problem_text):
|
|
value = _resolve_word_number(match.group(1), match.group(2))
|
|
if value is None:
|
|
continue
|
|
source_token = (
|
|
match.group(1)
|
|
if match.group(2) is None
|
|
else match.group(0).rsplit(maxsplit=1)[0]
|
|
)
|
|
found.append(
|
|
(
|
|
match.start(1),
|
|
Quantity(value=value, unit=_clean_unit(match.group(3)), source_token=source_token),
|
|
)
|
|
)
|
|
|
|
# 4. EX-5 — sentence-final bare numbers (empty unit).
|
|
for match in _FINAL_NUMBER_RE.finditer(problem_text):
|
|
if _claimed(match.start(1), claimed):
|
|
continue
|
|
quantity = _quantity(match.group(1), "")
|
|
if quantity is not None:
|
|
found.append((match.start(1), quantity))
|
|
|
|
found.sort(key=lambda item: item[0])
|
|
return tuple(quantity for _, quantity in found)
|