Commit graph

3 commits

Author SHA1 Message Date
Shay
268a754c00 fix: fold Gemini CGA/fraction_decrease compatibility into deck PR
Some checks failed
lane-shas / verify pinned lane SHAs (pull_request) Has been cancelled
smoke / smoke (-m "not quarantine") (pull_request) Has been cancelled
Port load-bearing fixes from Gemini worktree (Projects/core draft) onto
forgejo deck branch — reviewed and cleaned for invariants:

- fraction_decrease: optional contract with geometric proposal fallback;
  multi slash-fraction hazard refuse
- parse_and_solve: 9-decimal answer presentation boundary only (matches
  derivation verify uniqueness quantum; not algebra drift repair)
- proportional_decrease family asserts: serving_allowed after Gate A2k
- ADR-0184 S4b: re-pin equivalence corpus 937 → 1087

Not folded: local core-rs uv install (env-only), obsolete 3lang draft
from Gemini tree, double-rounding wrapper noise.
2026-07-08 20:00:51 -07:00
Shay
310aed9ff0
chore: Refactor CLI and Governance Anchors (#926)
* docs: consolidate governance anchors and clean up test registries

* refactor(cli): decompose cli into dedicated modules

* test: fix broken test baselines and formatting

* docs: add domain boundary READMEs for governance anchors

* test: update baseline for determination lane

* test: fix capability_pass expectation

* test: fix CORE_SHOWCASE_SKIP_BUDGET enforcement

* chore: cleanup CLI extraction and unreachable code
2026-07-03 12:34:56 -07:00
Shay
ad75a2c79a test(derivation): add ADR-0184 semantic replay equivalence harness
ADR-0184 S4b — the replay/provenance equivalence boundary, the
precondition for any semantic-primary path. After #685 the in-repo
legacy path delegates to the semantic boundary, so old-vs-new is
vacuous; the durable reference is a pinned canonical trace artifact
(evals/gsm8k_math/equivalence/v1/, 937 problems) whose authority is
the #684/#685 cross-tree differentials (documented provenance chain).

- state/provenance.py: replay_is_faithful — the replay bridge's law as
  a checkable structural property (no verify/pool import; covered by
  the #685 authority scan).
- equivalence/trace.py: canonical traces (worlds, candidates, order,
  multiplicity, classifications, resolutions) + authority_violations
  re-deriving the pool commit law from trace content.
- scripts/verify_semantic_equivalence.py: check / --update gate
  mirroring verify_lane_shas.py.
- tests/test_adr_0184_s4b_replay_equivalence.py: full-corpus live-vs-
  pinned equality, corpus-wide authority + faithfulness, and 17
  single-mutation non-vacuity proofs.

Pure additions; verify.py/pool.py untouched; no serving, CLAIMS,
metrics, or lane pins moved.
2026-06-10 14:35:22 -07:00