core-labs/core
core-labs/core
1
0
Fork 0
Code Issues Pull requests 1 Projects Releases Packages Wiki Activity

feat(trackb): S1–S4 symbolic SME, selector, pure-S1 coverage gain (Inc2) #86

Merged
core-labs merged 2 commits from feat/trackb-symbolic-sme-s2s4 into main 2026-07-20 03:12:52 +00:00
Conversation 0 Commits 2 Files changed 15 +2521 -393
core-labs commented 2026-07-20 02:49:32 +00:00
Owner
Copy link

Summary

Track B Increment 2 for ADR-0252: prove the structure-mapping path earns its keep (generalization ratio > 1, coverage gain, selector), without serving cutover.

Extends Increment 1 (generate/structure_mapping/) with:

  • Canonicals S1–S4 (pure-family gates; refuse supersets)
  • Overlapping-waves selector (rank, refuse ties/empty, structure-only ranking)
  • Generalized 3-gate wrong=0 solve corridor
  • SM-owned pure-S1 text extract (off-serving fallback when the live reader refuses) — recovers four real holdout cases: 0148, 0228, 0234, 0441

Bar results (command-backed)

Bar Result
Generalization ratio > 1 S1 = 9.0 (9 real holdout IDs / 1 template)
Coverage gain or organ retirement +4 holdout solves (organ 5 → trackb 9), wrong=0; no organ retirement
Selector S1/S2/S3 route correctly; empty refuses; transfer ≠ S1

Full tables + raw command output: docs/research/trackb-inc2-generalization-2026-07-20.md

Honest frontiers

  • S2–S4 holdout ratio = 0 — parser frontier: organ graphs on only 5/500 holdout cases (all S1). S2 proven on real public transfer graphs (engineering evidence, not holdout ratio).
  • S3/S4 emit blocked by multi-register scope (apply_rate / compare_additive out of Tier-2a) — map OK, refuse at cert (gates not weakened).
  • Serving reader unchanged; no organ retirement; sealed test untouched.

Test plan

  • uv run pytest tests/test_trackb_s1_structure_mapping.py tests/test_trackb_inc2_structure_mapping.py -q → 35 passed
  • uv run python scripts/measure_trackb_inc2.py --mode all → ratio 9.0, +4 coverage, selector green, wrong=0
  • uv run core test --suite smoke -q → 176 passed (~132s)

[Verification]: Smoke suite passed locally (~132s, 176 passed); trackb unit tests 35 passed; measure_trackb_inc2 all modes green.

STOP

Reviewer verifies against code (blindness, pure-family gates, real holdout IDs for ratio, three-gate wrong=0) before any serving cutover or next family expansion.

## Summary Track B **Increment 2** for ADR-0252: prove the structure-mapping path *earns its keep* (generalization ratio > 1, coverage gain, selector), without serving cutover. Extends Increment 1 (`generate/structure_mapping/`) with: - Canonicals **S1–S4** (pure-family gates; refuse supersets) - Overlapping-waves **selector** (rank, refuse ties/empty, structure-only ranking) - Generalized **3-gate wrong=0** solve corridor - **SM-owned pure-S1 text extract** (off-serving fallback when the live reader refuses) — recovers four real holdout cases: **0148, 0228, 0234, 0441** ## Bar results (command-backed) | Bar | Result | | --- | --- | | Generalization ratio > 1 | **S1 = 9.0** (9 real holdout IDs / 1 template) | | Coverage gain or organ retirement | **+4** holdout solves (organ 5 → trackb 9), `wrong=0`; no organ retirement | | Selector | S1/S2/S3 route correctly; empty refuses; transfer ≠ S1 | Full tables + raw command output: `docs/research/trackb-inc2-generalization-2026-07-20.md` ## Honest frontiers - **S2–S4 holdout ratio = 0** — parser frontier: organ graphs on only 5/500 holdout cases (all S1). S2 proven on real public transfer graphs (engineering evidence, not holdout ratio). - **S3/S4 emit** blocked by multi-register scope (`apply_rate` / `compare_additive` out of Tier-2a) — map OK, refuse at cert (gates not weakened). - **Serving reader unchanged**; no organ retirement; sealed test untouched. ## Test plan - [x] `uv run pytest tests/test_trackb_s1_structure_mapping.py tests/test_trackb_inc2_structure_mapping.py -q` → 35 passed - [x] `uv run python scripts/measure_trackb_inc2.py --mode all` → ratio 9.0, +4 coverage, selector green, wrong=0 - [x] `uv run core test --suite smoke -q` → 176 passed (~132s) `[Verification]: Smoke suite passed locally (~132s, 176 passed); trackb unit tests 35 passed; measure_trackb_inc2 all modes green.` ## STOP Reviewer verifies against code (blindness, pure-family gates, real holdout IDs for ratio, three-gate wrong=0) before any serving cutover or next family expansion.
core-labs added 1 commit 2026-07-20 02:49:36 +00:00 
Increment 2 for ADR-0252 Track B. Extends Increment 1's structure-mapping
slice with pure-family canonicals S2–S4, overlapping-waves selector, and a
structure-mapping-owned pure-S1 text extract that recovers four real
holdout cases the serving reader misses (0148, 0228, 0234, 0441).

Bar results (command-backed via scripts/measure_trackb_inc2.py):
- generalization ratio S1 = 9.0 (9 holdout cases / 1 template)
- coverage gain organ 5 → trackb 9, wrong=0
- selector routes S1/S2/S3; refuses empty; surface≠structure

Off-serving: no organ retirement, serving reader untouched. S2–S4 holdout
ratios are 0 (parser frontier). S3/S4 emit refuses at multi-register scope
without weakening the three-gate wrong=0 path.

[Verification]: Smoke suite passed locally (~132s, 176 passed);
trackb unit tests 35 passed; measure_trackb_inc2 all modes green.
core-labs added 1 commit 2026-07-20 03:04:39 +00:00 
The inverted-seed P2 regex treated optional "times" as optional, so
additive "8 more" on multi-entity zoo case 0361 extracted as pure S1 and
the three-gate corridor emitted 1.125 vs gold 114.0 (wrong≠0).

- Require explicit multiplicative surface (times as many / twice the …)
- Fail-closed purity gate: multi-clause markers, multi-numeric, multi-mult
- Regression tests for 0361 and bare "N more"
- measure --mode full-holdout-wrong0 over all 500 holdout cases

[Verification]: full-holdout emit_ok=9 wrong=0 refused=491; unit 37 passed
core-labs merged commit d8d62b8ea3 into main 2026-07-20 03:12:52 +00:00
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference: core-labs/core#86
No description provided.
Delete branch "feat/trackb-symbolic-sme-s2s4"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?