fix(scoring): Issue #262 — rename DQIC site-2 emission to valuation_output_anomalous + writer-parity for veto cohort UI#265
Merged
Conversation
…utput_anomalous + writer-parity for veto cohort UI Closes issue #262 (DQIC dual-surface emission inconsistency) per methodology-scientist Mode B verdict 2026-05-26 (APPROVED-AS-ANNOTATE, Path 3 = rename). The bug: data_quality_input_corruption emitted from TWO independent check sites with different trigger conditions: - Site 1 (risk_overlay.py:411) — INPUT-level corruption; risk_flags VETO - Site 2 (ensemble.py:545) — OUTPUT-level anomaly; valuation_warnings ANNOTATE Site 2's check is strictly broader. Universe scan 2026-05-23 cron #3: - 2 tickers fire BOTH (ERIE rank 69, BRK-B rank 223) - 4 tickers fire Site 1 ONLY (MTB/CPT/MRNA/HBAN) — UI explainability gap - 1 ticker fires Site 2 ONLY (NVR rank 267) — Top-5-safety gap if rose The bigger smell than NVR Top-5 risk is the UI gap: FairPriceCard.tsx:82 reads only valuation_warnings, so MTB/CPT/MRNA/HBAN render all-null fair-price with NO explanation chip. Path 3 fix: - Rename Site 2 (ensemble.py:533+545) → valuation_output_anomalous - Writer-parity in compute/main.py: when DQIC veto fires in risk_flags, ALSO emit valuation_output_anomalous to valuation_warnings (closes UI gap for veto-only cohort) - applicability.py SKIP_REASONS taxonomy gains valuation_output_anomalous (count 25 → 26); legacy data_quality_input_corruption retained for backward-compat on pre-rename JSON snapshots - sanity.py:83 IC-smoke exclusion ORs both identifiers - FairPriceCard.tsx:82 dataQualityIssue check ORs both identifiers Tests: - test_ensemble.py 4 assertions — assert new identifier on Site 2 emit - test_tier2_schema.py::test_B4_skip_reasons_count → 25 to 26 - test_sanity_smoke.py / test_recommendation.py unchanged (legacy-snapshot path verified via OR check; veto identifier unchanged) No schema bump (string-identifier rename only). SCHEMA_VERSION stays 0.10.5-phase4.5e. Triple lockstep unchanged. ZERO composite-rank impact — composite scores / risk_flags VETO identifiers / Top-5 rotation unchanged. Verification: ruff clean / schema_check clean / pytest 1216 passed (unchanged from PR #264 baseline). PHASE_STATUS_INFLIGHT.md side-file entry satisfies §Conventions "ship with every PR" lockstep per PR #237 convention. No CLAUDE.md / AGENTS.md substance change — rename doesn't introduce new invariant; methodology verdict documented in INFLIGHT entry. https://claude.ai/code/session_01JwntEE4PNAXSMkZxRA9BB4
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
Contributor
Pre-merge production simulation
Diff vs main
Main baseline: Top-10 movers (sorted by |Δcomposite_score|)
|
4 tasks
dackclup
added a commit
that referenced
this pull request
May 26, 2026
…Craft frontend (#266) Cuts the v1.3.0-phase4.5e release tag, closing the Phase 4.5e Form-4 insider-clustering ladder (PRs #167+#205+#222+#224+#238) and shipping the LedgerCraft frontend reskin (A1-A3+B1-B4+animation PRs 1-3+#244 polish+dark-mode tooltip fixes through PR #263) since v1.2.0-phase4.5 (6d414a9, 2026-05-17). Scope (3 files): - pyproject.toml — version 0.3.0 → 1.3.0 - docs/release-notes/v1.3.0-phase4.5e.md (NEW) — release body grouped by Form-4 cluster / data-quality / defense layer / frontend / methodology + agent infra / CI hygiene; ~800 words - PHASE_STATUS.md — Current state schema 0.10.4 → 0.10.5-phase4.5e, defense layer headline 32 → 33 declared flags, production-run pointer refreshed to 26423296287 Pre-flight ladder verified by release-captain (opus): - ruff clean - pytest 1216 passed (offline) - schema_check in sync at 0.10.5-phase4.5e - verify-production-output Section A-G + I-L PASS; Section H 1 known FAIL (orphan BK.json legacy snapshot, pre-existing) - frontend build verified via vercel-preview-auditor (sonnet) on main HEAD e6013ba — 506/506 routes compiled, types validated, runtime clean, 3-route UA probe PASS Defense scorecard: 7 active vetoes unchanged (altman_distress / sloan_accruals_top_decile / net_issuance_top_decile / non_reliance_filing / beneish_manipulation_veto / dechow_manipulation_veto / data_quality_input_corruption). Headline 32 → 33 declared boolean flags (adds multi_class_aggregate_shares_suspected per PR #264; PR #265 DQIC rename is identifier-shape, not new flag). Production output: metadata.json reports 0.10.4-phase4.5e from cron #4 (2026-05-26T01:12); next weekday cron Wed 2026-05-27 22:00 UTC re-renders at full 0.10.5-phase4.5e semantics. Tag is anchored to code, not last committed snapshot per release-tag SKILL.md §Gotchas. CVE baseline 25 → 15 open (0C/6H/7M/2L); all 15 are next@14.x SSR advisories with zero exploitability on static-export. Post-merge: tag command + GitHub Release creation require explicit user authorization per CLAUDE.md §Executing actions with care. https://claude.ai/code/session_01JwntEE4PNAXSMkZxRA9BB4 Co-authored-by: Claude <noreply@anthropic.com>
7 tasks
dackclup
added a commit
that referenced
this pull request
May 27, 2026
…AUDE.md (#271) Refactors a user-shared research report (Master Prompt + 6 phase sub-prompts + CLAUDE.md template) into the existing doc surface without creating a new .claude/skills/agentic-6-phase/ skill. The report's underlying logic is already implemented in the 18 subagents + CLAUDE.md §Auto-routing; what was genuinely missing was a 6-phase mapping table a new session can scan in < 30 sec on top of the 9 phases. Scope (2 substance files + 1 INFLIGHT entry): - WORKFLOW.md — new section "Agentic 6-Phase Cadence" between §"Tools You'll Use Daily" and §"Phase Overview". Mapping table (Step × Fire trigger × Subagent(s) × Done when) over Planning → Code Gen → Integration → Test → Deploy → Monitor + 5 cadence invariants. Reuses the 18 standing subagents — no new agent files. Session-start protocol cites schema 0.10.5-phase4.5e (PRs #264 + #265; cron #4 still at 0.10.4, next cron Wed 2026-05-27 re-renders at 0.10.5), defense layer 33 declared = 7 vetoes + 26 annotates, tag v1.3.0-phase4.5e, CVE baseline 15 open (0C / 6H / 7M / 2L) after PR #194 patch + PR #226 triage. - CLAUDE.md — new §Conventions bullet "Session-start phase identification" (~5 lines) pointing readers at PHASE_STATUS.md §"Current state" + WORKFLOW.md §"Agentic 6-Phase Cadence" using the standing 18 subagents. - PHASE_STATUS_INFLIGHT.md — new in-flight entry per PR #237 side-file lockstep convention. Out of scope (deliberately NOT done per user direction 2026-05-27): - NO .claude/skills/agentic-6-phase/ — overhead exceeds benefit - NO Master Prompt / phase sub-prompts copied into the repo - NO edits to any of the 18 subagent files under .claude/agents/ - NO AGENTS.md substance edit — the cadence is Claude-Code-subagent- specific; cross-tool agents would route differently. INFLIGHT entry satisfies §Conventions "ship with every PR" lockstep. docs-reviewer verdict (2026-05-27, agent id a2c87ed3679f55fe5): NEEDS-CROSS-REF-FIX — both items applied in this commit: 1. CVE attribution: "after PR #226 triage" → "after PR #194 patch + PR #226 triage" (PR #194 closed the 10 advisories; PR #226 documented the resulting state) 2. Step 4 fire-trigger col: "Sections A-J" → "Sections A-L" (Section L added by PR #221 OSAP proxy invariant; internal match with the same row's Done-when col) All else passes: 4 cited numbers, 18 agent names, 3 cross-refs, token budget (WORKFLOW ≤ 1 page, CLAUDE ≤ 5 lines), Rule 16 + Rule 18 no contradiction. Pre-existing SKILL.md schema-version table gap (rows for 0.10.5-phase4.5e PR #264 + valuation_output_anomalous rename PR #265 missing) escalated to schema-sentinel as separate doc-only PR per docs-reviewer recommendation — not blocking on this scope. Verification: - ruff check . — N/A (no Python) - python -m compute.output.schema_check — N/A (no schemas) - pytest tests/ -m "not network" — N/A (no test surface) - docs-reviewer subagent — PASS after the 2 fixes above Co-authored-by: Claude <noreply@anthropic.com>
dackclup
pushed a commit
that referenced
this pull request
May 28, 2026
Addresses 3 findings from docs-reviewer (sonnet) substance review of the housekeeping PR (commit e060cb9): 1. PHASE_STATUS.md:98 — "v1.3.0 release tag" entry in §Next deliverables was stale (v1.3.0 shipped 2026-05-26, v1.4.0 shipped 2026-05-27). Replaced with forward-look for v1.5.0 gated on Phase 4.5e PR 5 + Issue #67 sector-CoE flip. 2. WORKFLOW.md:730 — Phase 4.5 historical task list trailing clause "v1.3.0 target pending release-captain ladder (LedgerCraft A-B series + this doc-refresh)" was stale. Replaced with the actual landing dates + SHAs for both v1.3.0 (5db3b97) and v1.4.0 (bbca9ca). 3. docs/METHODOLOGY.md — annotate-only flag count out of sync with declared flags: - Line 16: "21 annotate-only flags" → "23 annotate-only flags" - Line 164: section heading "(21)" → "(23)" - Added 2 new bullets at end of §Annotate-only flags section: - `multi_class_aggregate_shares_suspected` (Issue #261 PR-A, PR #264) — CIK-collision detector; identity-equation check per Damodaran 2019 Ch. 16; no academic prior (data-quality) - `valuation_output_anomalous` (Issue #262, PR #265) — Site-2 output-anomaly detector renamed from `data_quality_input_corruption`; no academic prior (data-quality); semantic distinction from Site-1 input-corruption veto documented inline Both new METHODOLOGY entries are data-quality detectors with no new academic claim — `multi_class_aggregate_shares_suspected` cites the existing Damodaran 2019 anchor (already in the file) for the per-class market-cap identity equation, and `valuation_output_anomalous` is the renamed Site-2 emission of the existing `data_quality_input_corruption` defense. Neither requires methodology-scientist verdict per the "Internal — data-quality" pattern shared with `goodwill_heavy` / `data_quality_input_corruption` Site-1 already in the section.
dackclup
added a commit
that referenced
this pull request
May 28, 2026
…v1.4.0 (#286) * chore(docs): housekeeping PR-B — drain INFLIGHT + bump pointers post-v1.4.0 Phase B post-tag housekeeping. Drains 7 stale `(in flight, ...)` markers from PHASE_STATUS_INFLIGHT.md (PRs #269, #267, #271, #280, #281, #282, #285) to `(merged YYYY-MM-DD, <SHA>)`, and bumps stale schema/tag pointers across CLAUDE.md / PHASE_STATUS.md / SKILL.md / WORKFLOW.md to reflect the v1.4.0-phase4.6 release at `bbca9cac`. Changes: - CLAUDE.md §Phase status — schema `0.10.2-phase4.5e` → `0.10.7-phase4.6`, tag `v1.3.0-phase4.5e` → `v1.4.0-phase4.6` (2026-05-27, `bbca9cac`), "Recently merged" list refreshed from PRs #147-#154 → PRs #264-#285 - PHASE_STATUS.md §Current state — mirrored pointer bump; production-run pointer → `559c5269` (cron-#5 2026-05-27 chore commit); "Recently merged" prepended with 22 entries since v1.3.0, legacy list relabeled as "Earlier" - SKILL.md schema-version table — 3 new rows: `0.10.7-phase4.6` (PR #283 release), `0.10.6-phase4.5e` (PR #269 GOOG/GOOGL per-class XBRL fix), `0.10.5-phase4.5e` (PR #264 multi-class CIK detector) - WORKFLOW.md §Agentic 6-Phase Cadence Session-start protocol — pointer block bumped to current state - PHASE_STATUS_INFLIGHT.md — 7 stale markers drained + new entry for this PR appended at end AGENTS.md substance untouched per the existing delegation pattern (line 372-375: "Canonical 'current state' lives in CLAUDE.md §Phase status. Schema-version history table is in SKILL.md."). Cross-tool agents reading state pull from CLAUDE.md as the source of truth. Doc-only PR — `ruff` / `schema_check` trivially pass; no compute / schema / scoring / valuation / frontend / Python / TS code change. * docs(review-fix): docs-reviewer NEEDS-CROSS-REF-FIX — 3 items Addresses 3 findings from docs-reviewer (sonnet) substance review of the housekeeping PR (commit e060cb9): 1. PHASE_STATUS.md:98 — "v1.3.0 release tag" entry in §Next deliverables was stale (v1.3.0 shipped 2026-05-26, v1.4.0 shipped 2026-05-27). Replaced with forward-look for v1.5.0 gated on Phase 4.5e PR 5 + Issue #67 sector-CoE flip. 2. WORKFLOW.md:730 — Phase 4.5 historical task list trailing clause "v1.3.0 target pending release-captain ladder (LedgerCraft A-B series + this doc-refresh)" was stale. Replaced with the actual landing dates + SHAs for both v1.3.0 (5db3b97) and v1.4.0 (bbca9ca). 3. docs/METHODOLOGY.md — annotate-only flag count out of sync with declared flags: - Line 16: "21 annotate-only flags" → "23 annotate-only flags" - Line 164: section heading "(21)" → "(23)" - Added 2 new bullets at end of §Annotate-only flags section: - `multi_class_aggregate_shares_suspected` (Issue #261 PR-A, PR #264) — CIK-collision detector; identity-equation check per Damodaran 2019 Ch. 16; no academic prior (data-quality) - `valuation_output_anomalous` (Issue #262, PR #265) — Site-2 output-anomaly detector renamed from `data_quality_input_corruption`; no academic prior (data-quality); semantic distinction from Site-1 input-corruption veto documented inline Both new METHODOLOGY entries are data-quality detectors with no new academic claim — `multi_class_aggregate_shares_suspected` cites the existing Damodaran 2019 anchor (already in the file) for the per-class market-cap identity equation, and `valuation_output_anomalous` is the renamed Site-2 emission of the existing `data_quality_input_corruption` defense. Neither requires methodology-scientist verdict per the "Internal — data-quality" pattern shared with `goodwill_heavy` / `data_quality_input_corruption` Site-1 already in the section. --------- Co-authored-by: Claude <noreply@anthropic.com>
This was referenced May 28, 2026
dackclup
added a commit
that referenced
this pull request
May 28, 2026
…(NVR FP) (#293) * fix(valuation): Issue #289 — retire Site-2 DQIC ceiling per Option C Closes Issue #289 — NVR ($458 EPS, $6,098 price, ~2.7M shares) tripped the $10,000 absolute-dollar Site-2 ceiling (`multiples_pe ≈ 22× × $458.86 ≈ $10,094`) and got all 6 valuation methods nulled despite legitimate inputs. `/stock/NVR` rendered empty fair-price section on cron #69. Per methodology-scientist Mode B verdict 2026-05-28 = Option C (LITERATURE-ANCHORED on Penman 2013 §7.4 + Damodaran 2019 Ch. 18 + Huber 1981 §1.4): - Site-2 was structurally redundant with Defense #4 (`extreme_*_estimate`) + Issue #177 `extreme_estimate_majority` (Huber breakdown layer) - Site-1 input-level check in `risk_overlay.py` already catches the upstream units-bug class via TBVPS/revenue/NI patterns - Cohort impact on cron #69: Option A ($50K) = 0 newly flagged; Option B (5× ratio) = 12 NEWLY NULLED tickers (regression); Option C = 0 newly missed - PPV on cron #69 for Site-2: 0/1 = 0% (NVR was the only firing, FP) Changes (5 files): - compute/valuation/ensemble.py:450 — Site-2 trigger DELETED (`if _has_corrupt_input(methods): return _data_quality_corrupt_result`). `_has_corrupt_input` + `_data_quality_corrupt_result` retained below as dead code for one cycle per methodology verdict. - compute/config.py:121-138 — `FAIR_PRICE_DATA_QUALITY_CEILING = 10000.0` KEPT ACTIVE. The constant is shared with Site-1 (`risk_overlay.py:149` TBVPS > $10K input check); commenting it out would break the input-corruption veto. Docstring rewritten to confirm Site-1 active use + document Site-2 retirement. - compute/main.py:1495-1499 — Writer-parity emit UNCHANGED. When `data_quality_input_corruption` veto fires in `risk_flags`, `valuation_output_anomalous` is still appended to `valuation_warnings` so the UI explanation chip persists for Site-1 cohort (MTB / CPT / MRNA / HBAN per PR #265). - docs/METHODOLOGY.md — `valuation_output_anomalous` annotate description re-anchored: "Site-2 retired; emit remains via writer-parity from compute/main.py on Site-1 cohort". Cites Penman 2013 §7.4 + Damodaran 2019 Ch. 18 + Huber 1981 §1.4 + 0/1 PPV empirical justification. - tests/test_valuation/test_ensemble.py — 1 test rewritten as retirement guard (`test_site2_data_quality_guard_retired_post_issue_289`) + 3 new Section L tests authored by test-engineer (NVR regression + dead code retention + constant preservation guards). Impact (display-only): - /stock/NVR now renders fair-price section with median of surviving methods - NVR composite + rank UNCHANGED (composite = 50.99, rank #252) - Rule 16 + Top-5 rotation UNAFFECTED - Site-1 input veto unchanged (4 cohort tickers keep VETO + UI chip) - Defense layer count unchanged at 33 declared Verification: - `ruff check .` — PASS - `python -m pytest tests/test_valuation/test_ensemble.py -q -m "not network"` — 54 passed - `python -m pytest tests/ -q -m "not network" --ignore=tests/test_validation` — 1203 passed Deferred (NOT in this PR): - Dead-code removal of `_has_corrupt_input` + `_data_quality_corrupt_result` after ≥ 1 cron of clean operation confirms no regression - Site-1 input threshold recalibration — Q3 2026-08-19 cohort audit slot Mid-PR correction: initial commenting-out of `FAIR_PRICE_DATA_QUALITY_CEILING` was reverted after grep revealed Site-1 shares the constant. The retirement applies to the Site-2 trigger only, not to the shared infrastructure constant. * test(valuation): test-engineer follow-up edits to Section L tests Post-PR-#293-push edits from `test-engineer` (sonnet) that landed after the initial 4b5614e commit. Polishes the 3 new Section L Issue #289 retirement-guard tests: - L1 NVR fair-price methods no longer nulled — synthetic NVR-shape fixture cleanup + assertion tightening - L2 dead-code function-retention guard — confirm `_has_corrupt_input` + `_data_quality_corrupt_result` still callable post-Issue-#289 - L3 rewritten as `test_L3_site2_ceiling_not_invoked_when_method_value_exceeds_10k` — uses minimal synthetic methods dict (multiples_pe=$10,001 against current_price=$3,000 = 3.33× within Defense #4 5× band) so the retirement-guard doesn't depend on peer-panel infrastructure; confirms neither `extreme_multiples_pe_estimate` NOR `valuation_output_anomalous` appears in ensemble warnings post-fix Net change: +59 / -46 lines on the same file; all 54 tests in tests/test_valuation/test_ensemble.py pass post-fix. No production code change. --------- Co-authored-by: Claude <noreply@anthropic.com>
6 tasks
dackclup
added a commit
that referenced
this pull request
May 28, 2026
PR #293 (95e638b, merged 2026-05-28 05:20 UTC) retired the Site-2 output-level data-quality ceiling per methodology-scientist Mode B Option C verdict (Penman 2013 §7.4 + Damodaran 2019 Ch. 18 + Huber 1981 §1.4). PR #293 deleted the call site at compute/valuation/ ensemble.py Step 4.5 but RETAINED the 2 helper functions as dead code for "one cycle hold" — explicit retention guard pinned the deferred state. This PR closes that contract. Cron Run #71 (368dccd, 2026-05-28 08:44 UTC) confirmed clean operation: - NVR fair-price section renders correctly (was empty pre-PR-#293) - valuation_output_anomalous cohort dropped 5 → 4 (NVR removed) - No regression on Site-1 veto cohort (MTB / CPT / MRNA / HBAN per PR #265 writer-parity emit still firing) Scope (8 files, net −41 lines): - compute/valuation/ensemble.py — REMOVED 2 dead functions: - _has_corrupt_input(methods) -> bool (13 lines) - _data_quality_corrupt_result(methods) -> EnsembleResult (43 lines) Step 4.5 comment block at lines 449-479 refreshed: "kept as dead code for one cycle" → "removed in this PR after cron Run #71 confirmed clean". - tests/test_valuation/test_ensemble.py — REMOVED 2 imports + 3 active tests that exercised the removed functions: - test_data_quality_sanity_guard_triggers_on_extreme_method_value - test_data_quality_guard_boundary_exactly_at_ceiling - test_data_quality_guard_skipped_methods_dont_trigger The one-cycle retention guard test_L2_dead_code_functions_still_ callable_after_site2_deletion REPLACED with new test_L2_dead_code_functions_removed_post_one_cycle that pins the removal via `not hasattr(_ensemble, "_has_corrupt_input")` + `not hasattr(_ensemble, "_data_quality_corrupt_result")` — so a future accidental re-introduction surfaces as a clear "Issue #289 retirement reverted" failure. The 2 surviving tests (test_site2_data_quality_guard_retired_post_issue_289 + test_L3_site2_ceiling_not_invoked_for_high_share_price_ticker) keep verifying POST-RETIREMENT invariants on the NVR cohort. - compute/config.py:127-148 — FAIR_PRICE_DATA_QUALITY_CEILING STAYS ACTIVE (Site-1 in compute/scoring/risk_overlay.py shares the constant). Comment block updated to note "dead-code helpers REMOVED in the PR #293 follow-up after cron Run #71 confirmed clean". - compute/valuation/applicability.py:65-78 — reference to _data_quality_corrupt_result replaced with reference to the writer-parity emit at compute/main.py on the Site-1 veto cohort. - tests/test_scoring/test_risk_overlay.py:520-528 — test_D4_data_quality_corruption_fires_at_boundary_strict docstring updated: "Mirrors _has_corrupt_input's strict inequality" → "Site-1 (here) is the canonical input-corruption guard. Site-2 was retired per Issue #289 Option C; the strict `>` invariant lives ONLY here now." Test logic UNCHANGED. - CLAUDE.md §Phase status — drained "(in flight this PR — Issue #67 follow-up)" wording (PR #300 merged); added in-flight note for this PR's Site-2 dead-code removal. - AGENTS.md open-issues #289 entry — appended "dead-code helpers REMOVED in the PR #293 follow-up after cron Run #71 confirmed the retirement empirically". - PHASE_STATUS_INFLIGHT.md — full in-flight entry appended per PR #237 side-file convention. Verification ladder: - ruff check . PASS - python -m compute.output.schema_check PASS (no schema touched) - python -m pytest tests/test_valuation/test_ensemble.py tests/test_scoring/test_risk_overlay.py tests/test_config.py -q 120 PASS (was 123 — net -3 as planned: 3 deleted tests + 1 modified L2 guard) - grep -rn "_has_corrupt_input|_data_quality_corrupt_result" compute/ tests/ only legitimate references remain (retirement comments + new removal-guard assertion strings) Hard constraints honored: - No scoring formula change · No Rule 16 / Top-5 violation - Schema version UNCHANGED at 0.10.10-phase4.6 - Site-1 input-corruption veto path UNCHANGED - FAIR_PRICE_DATA_QUALITY_CEILING constant retained (Site-1 shared) - Writer-parity valuation_output_anomalous emit at compute/main.py UNCHANGED — UI explanation chip continues for Site-1 cohort - test_L3 PR #293 end-to-end guard UNCHANGED — keeps verifying NVR cohort produces non-null median Closes the PR #293 retirement contract. https://claude.ai/code/session_01AGU8d6pm4u2fQQ5cebg9qa Co-authored-by: Claude <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Closes issue #262 (DQIC dual-surface emission inconsistency) per methodology-scientist Mode B verdict 2026-05-26 (
APPROVED-AS-ANNOTATE, Path 3 = rename).The dual-surface bug
data_quality_input_corruptionwas emitted from TWO independent check sites with DIFFERENT trigger conditions:compute/scoring/risk_overlay.py:411risk_flags(VETO)compute/valuation/ensemble.py:545valuation_warnings(ANNOTATE)Site 2's check is strictly broader than Site 1's.
Universe scan (2026-05-23 cron #3, methodology-scientist walk)
risk_flags+valuation_warningsThe bigger smell than the NVR Top-5 risk is the UI explainability gap:
FairPriceCard.tsx:82reads onlyvaluation_warnings, so MTB/CPT/MRNA/HBAN render the all-null fair-price ensemble with NO explanation chip.Path 3 fix (this PR)
1. Rename Site 2 emission
compute/valuation/ensemble.py::_data_quality_corrupt_resultnow emitsvaluation_output_anomalous:reasonfield (line 533)valuation_warningslist (line 545)Semantically distinct from the input-level
data_quality_input_corruption: "a method produced an absurd output despite plausible inputs" is NOT categorical evidence of input untrust — could be residual input bug Site-1 missed, legitimately extreme RIM on a goodwill-heavy filer, OR a method-formula edge case. Annotate-only is the correct Rule 16 surface.2. Writer-parity emit in
compute/main.pyWhen
data_quality_input_corruptionis inrisk_flags.get(ticker, [])(the VETO surface), ALSO appendvaluation_output_anomaloustovaluation_warningsfor UI parity. Closes the veto-only-cohort UI explainability gap (MTB/CPT/MRNA/HBAN now gain the UI chip).3. Consumer updates
compute/valuation/applicability.py—SKIP_REASONStaxonomy gainsvaluation_output_anomalous; legacydata_quality_input_corruptionretained for backward-compat on pre-rename JSON snapshots. Count 25 → 26.compute/scoring/sanity.py:83—compute_mos_trailing_icIC-smoke exclusion ORs both identifiers.frontend/components/FairPriceCard.tsx:82—dataQualityIssueflag check ORs both identifiers (backward-compat for pre-rename snapshots still on the static site between cron-feat(phase-2): SEC EDGAR fundamentals + per-stock detail pages #4 and cron-feat(phase-3a): pillar feature modules — 30+ classical metrics #5).4. Tests
tests/test_valuation/test_ensemble.py— 4 assertions updated to expect new identifiertests/test_output/test_tier2_schema.py::test_B4_skip_reasons_count_is_25→_is_26tests/test_scoring/test_sanity_smoke.py(2 sites) — unchanged; legacy-snapshot path verified via the OR checktests/test_scoring/test_recommendation.py(2 sites) — unchanged; veto identifier (risk_flags) never renamedNo schema bump
String-identifier rename only — no new
Metadata/StockDetailfield.SCHEMA_VERSIONstays at0.10.5-phase4.5e. Triple lockstep unchanged.ZERO composite-rank impact
Composite scores /
risk_flagsVETO identifiers / Top-5 rotation unchanged. The only behavioral effect is which identifier appears invaluation_warnings(a display field) and gains the writer-parity for the veto cohort UI.Verification ladder
ruff check .— All checks passedpython -m compute.output.schema_check— clean (no schema change)pytest tests/ -m "not network"— 1216 passed, 7 skipped, 24 deselectedDeferred follow-ups
valuation_output_anomalousfiring list, decide whether to retire the legacydata_quality_input_corruptionidentifier from the taxonomy after ≥ 2 crons of clean rename adoption (methodology Q6)valuation_output_anomalousvs existingextreme_*_estimateannotates (methodology Q3) — confirms the new identifier is independent enough to justify the taxonomy slotLockstep
PHASE_STATUS_INFLIGHT.mdentry appended per PR docs(workflow): adopt PHASE_STATUS_INFLIGHT.md side-file (structural fix for parallel-PR collision) #237 side-file convention; PR feat(scoring): Issue #261 PR-A — multi_class_aggregate_shares_suspected annotate (CIK-collision detector) #264 entry moved into the historical record alongsideTest plan
/stock/MTB//stock/CPT//stock/HBANto verify the UI annotate appears AFTER cron-feat(phase-3a): pillar feature modules — 30+ classical metrics #5 lands (will be missing pre-cron until next compute writes the new identifier)https://claude.ai/code/session_01JwntEE4PNAXSMkZxRA9BB4
Generated by Claude Code