fix(,#19607): Planners-9 HTN -- C.4 stochasticite aries, prose 'MEME plan' etait fausse by jsboige · Pull Request #19675 · jsboige/CoursIA

jsboige · 2026-10-07T07:05:31Z

Diagnostic derive

Cause C.4 (e) : stochasticite non seedee du moteur aries (up_aries 0.5.0).

OneshotPlanner selectionne aries automatiquement a partir du problem_kind HTN. Le reseau de taches ajoute les deux deliver (pkg1 vers store, pkg2 vers office) sans set_ordered, et aries est un solveur satisficing : 3 plans valides de meme longueur (7 actions), tirages non-deterministes.

4 executions Linux (CPython 3.11, instance identique) :

Plan	Sequence	Tirages
A	`load pkg1 -> drive depot->warehouse -> load pkg2 -> drive warehouse->office -> unload pkg2 -> drive office->store -> unload pkg1`	1/4
B	`load pkg1 -> drive depot->warehouse -> load pkg2 -> drive warehouse->store -> unload pkg1 -> drive store->office -> unload pkg2`	2/4
C	`load pkg1 -> drive depot->store -> unload pkg1 -> drive store->warehouse -> load pkg2 -> drive warehouse->office -> unload pkg2`	1/4

Le C# commit (cellule 22 de Planners-9-HTN-Csharp.ipynb) montre le plan A. Le Python commit (Planners-9-HTN.ipynb §5.2) montre le plan C. Deux tirages distincts du meme solveur.

Stable : les 3 premieres actions (repositionnement via goto + chargement conjoint).
Variable : l'ordre des 2 dernieres livraisons (les deux deliver non-ordonnees).

Volet prose -- cellule C# 21 re-alignee

L'ancienne cellule C# 21 affirmait : « renvoie un plan de 7 actions -- le meme plan que le jumeau Python » et decrivait le plan A comme etant la strategie unique. Le jumeau Python (cellule 32) avait deja la nuance (« aries etant un solveur satisficing, l'ordre exact des deux dernieres livraisons peut varier d'une execution a l'autre ; le coeur repositionnement + chargement conjoint, lui, est stable »). Realignement du C# sur cette nuance :

Les 3 plans valides sont nommes (A, B, C) avec leurs tirages respectifs.
Ce qui est stable (3 premieres actions) vs ce qui varie (ordre des 2 livraisons) est explicite.
La citation du plan A reste presente (c'est le tirage committe dans la cellule 22), mais n'est plus presente comme etant le seul.

Volet registre -- `planners-9-htn.yaml` corrige

bridge_verdict_reason affirmait a tort « le pont resout et renvoie le MEME plan de 7 actions que le twin Python ». Remplace par « un plan parmi 3 » avec les 3 plans A/B/C et la cause C.4 (e). Nouvelle entree known_differences « Stochasticite HTN (c.220, #19607) » pour tracabilite.

Pourquoi MD-only (pas de re-exec kernel)

Aucun output de cellule code n'est modifie (cellule 22 = sorties plan A, valides pour un tirage). Re-executer deriverait le kernel et remplacerait A par B ou C (cf Tell c.18570 : preferer MD-only si re-exec drifterait kernel). C.2 ne s'applique pas : aucune cellule code source touchee.

Acceptance (3 cellules-types verifiees)

Cellule C# 21 : passage de « le meme plan que le jumeau Python » a « Le plan n'est pas unique -- 3 plans valides », plans A/B/C nommes.
Cellule Python 32 : inchangee (la nuance y etait deja).
Registre planners-9-htn.yaml : bridge_verdict_reason realigne, known_differences enrichi.

Verdict C.4 : CAUSE_FIXED -- cause identifiee, prose alignee, registre corrige, aucune cellule code touchee.

Lien

Voir #19607

Grain: MED/notebook-dotnet -- lane myia-ai-01:CoursIA-2 -- prev: LIGHT/docs #19649

🤖 Generated with Claude Code

…tait faux Aries etant un solveur *satisficing* et le reseau 'deliver' n'ordonnant pas les deux taches, j'observe 3 plans valides de meme longueur (4 executions Linux) : A (C# commit), B (majoritaire), C (Python commit). Cause C.4 (e) : stochasticite non seedee du moteur, pas une regression. Le registre 'bridge_verdict_reason' de planners-9-htn.yaml affirmait a tort 'le pont resout et renvoie le MEME plan que le twin Python'. Le jumeau Python (cell 32) avait deja la nuance ; le C# (cell 21) sur- promettait. Realignement : - Planners-9-HTN-Csharp.ipynb cell 21 : les 3 plans sont nommes (A/B/C) avec leurs tirages respectifs, ce qui est stable (3 premieres actions : repositionnement + chargement conjoint) vs ce qui varie (ordre des 2 dernieres livraisons) explicite. - planners-9-htn.yaml bridge_verdict_reason : 'un plan parmi 3' remplace 'MEME plan', avec les 3 plans A/B/C et la cause C.4 (e). - nouvelle entree known_differences : stochasticite HTN documentee. Aucun output de cellule code modifie (cell 22 = sorties plan A, valides pour un tirage ; re-exec derait kernel -- MD-only preserve). Aucun re-execution kernel necessaire (C.2 ne s'applique pas : pas de cellule code source touchee). Lane myia-ai-01:CoursIA-2. Grain: MED/notebook-dotnet -- lane myia-ai-01:CoursIA-2 -- prev: LIGHT/docs #19649 Co-Authored-By: Claude Haiku 4.5 (1M context) <noreply@anthropic.com>

github-actions · 2026-10-07T07:27:10Z

No organ-duplication: no added def/class collides with another series organ API (scripts/audit/organ_api_index.yaml).

Detector: python scripts/audit/detect_organ_duplication.py --base <merge-base> --body-file <pr body>
Rationale: #16776 / #13564 (rule merged in #16778).

github-actions · 2026-10-07T07:27:13Z

✅ No prose/output mismatch detected in the notebooks this PR changed.

Scope = notebooks CHANGED in this PR, not the whole corpus. Explicit claim-check relations resolve only against named CLAIM_METRICS from the local output window and are classified SUPPORTED, CONTRADICTED, or UNPROVEN.
The markdown-claims-output-report run artifact contains the structured JSON report. See python scripts/check_markdown_claims_output.py --help for re-running locally.
Detector rationale: c.290 / c.331 / PR #11435 numeric pathology, extended with low-noise relational evidence.

github-actions · 2026-10-07T07:27:16Z

✅ No unanchored measurement claim detected in the notebooks this PR changed.

Scope = notebooks CHANGED in this PR, not the whole corpus. The stale-claim-report run artifact holds the structured JSON.
Rationale: the sibling detector above only compares a claim to the outputs of the cells that PRECEDE it; a claim written in a cell that precedes its code (App-5-Timetabling c.2/c.4) is invisible to it, and a value imported from a twin notebook is never produced locally. See python scripts/check_stale_claims.py --help.

github-actions · 2026-10-07T07:27:18Z

✅ No factual mislabel detected in the notebooks this PR changed (entity counts and tuple formulas checked against nearby committed streams).

Scope = notebooks CHANGED in this PR, not the whole corpus. The factual-mislabel-report run artifact holds the structured JSON.
Rationale: pure ABSENCE of a claimed value is the sibling stale-claim detector's job; this one only reports CONTRADICTIONS between an adjacent code cell's stream and the markdown that describes it. See python scripts/check_factual_mislabel.py --help.

github-actions · 2026-10-07T07:29:14Z

G-VAR-2/3 GENRE signals (advisory, non bloquant, #10020).
La lane `myia-ai-01:CoursIA-2` voit ces signaux actifs sur les mergees du jour (UTC 2026-10-07) :

TIER-INFLATION : declared LIGHT << effective LIGHT-genre (tally : declared=0 genre=2 cap=1)
CAP-EXCEEDED-BY-GENRE : light_genre > cap partage G-VAR-2 (tally : declared=0 genre=2 cap=1)
NOTE ([variation] Le label est lane-agregat mais PR-attache : le merge-gate peut HOLD le grain de CONTENU qui remedie au motif #10341) : la PR courante est de classe CONTENU (non LIGHT-genre) et ne contribue pas au motif ci-dessus -- les labels agregees ne sont PAS poses sur cette PR (le merge-gate ne doit pas la HOLD pour ce motif ; le coupable est parmi les grains META de la lane).

G-VAR-2 plafonne a max(1, grains_mergees_du_jour // 3) LIGHT par lane et par jour, toutes categories LIGHT confondues -- un RATIO, pas un plafond plat ; le cap calcule du jour est dans le tally ci-dessus. G-VAR-3 interdit deux genres LIGHT consecutifs. Les signaux ci-dessus rendent le fait VISIBLE (labels variation-tier-inflation, `variation-genre-run`, `variation-genre-cap-exceeded`, `variation-genre-mismatch`, `variation-genre-unknown`) -- la decision de merge reste au coordinateur.

github-actions · 2026-10-07T07:32:54Z

Golden-Set Execution (H.7 P3)

✅ 9/9 notebooks passed (certified reproducible)

Notebook	Status	Time
2.1-Workflow-ML.ipynb	✅ SUCCESS	5.2s
2.2-Descente-de-gradient.ipynb	✅ SUCCESS	6.4s
2.3-Regression-lineaire-logistique.ipynb	✅ SUCCESS	8.2s
2.4-Arbres-Forets-Ensembles.ipynb	✅ SUCCESS	8.8s
Search-01-StateSpace.ipynb	✅ SUCCESS	6.8s
SL-1-LogicalLearning.ipynb	✅ SUCCESS	4.5s
RL-04-Bandits-Manchots-Python.ipynb	✅ SUCCESS	50.0s
GameTheory-04c-NashExistence-Python.ipynb	✅ SUCCESS	20.0s
GameTheory-13d-Optimistic-CFR-Python.ipynb	✅ SUCCESS	22.8s

Pinned lockfile: scripts/notebook_tools/golden_set.lock.txt (H.7 P3, axe A #4208)

github-actions · 2026-10-07T08:10:33Z

Notebook outputs-required (H.4 schema): PASS (every code cell carries an `outputs: list`)

github-actions · 2026-10-07T08:13:57Z

Notebook PR Validation: PASS

Notebooks checked: 1
Code cells validated: 10
Result: All passed

Checks: H.1 (no errors), H.3 (execution_count), C.1 (no banned patterns)
Non-Python kernels (.NET/Lean): C.1 + errors only (execution_count advisory)
QuantConnect notebooks: C.1 + errors only (require QC Cloud for execution)

Le commit 6c8f557 a modifie le C# twin (cell 21, plans A/B/C expliques) suite au diagnostic 'MEME plan etait faux' de la stochasticite aries. Le content_csharp_sha a drifte eacb90b6 -> 230ff7af ; le Python twin n'a pas bouge (cell 32 avait deja la nuance). Paraphite-preservant : axe de parite semantic preserve, la stochasticite du moteur est documentee dans known_differences de planners-9-htn.yaml. Audit date: 2026-10-07, lane myia-ai-01:CoursIA-2. Refs #19607, #19675.

jsboige · 2026-10-07T15:54:53Z

[ADJOINT PREFLIGHT]
schema: 1
lane: myia-po-2026:CoursIA-3
pr: 19675
head: 27fb0b3
complete: true
body: read
comments-reviewed: 8
reviews-reviewed: 0
threads-reviewed: 0
threads-unresolved: 0
surfaces-sha256: 0140c8cfa534c4c906b9c971117a40e46adf4511e7c11c330b62199400549fc9
diff-files: 3
diff-additions: 20
diff-deletions: 14
checks: latest-wins-green_OR_BLOCKED
b0: clear_OR_blocked
scope: pass_OR_fail
domain: pass_OR_not-applicable_OR_fail
verdict: BLOCKED
organ: check_adjoint_prevalidation.py
organ-command: python scripts/check_adjoint_prevalidation.py --derive-verdict 19675 19675
organ-rc: 3
[/ADJOINT PREFLIGHT]

jsboige · 2026-10-08T01:47:17Z

[ADJOINT PREFLIGHT]
schema: 1
lane: myia-po-2026:CoursIA-3
pr: 19675
head: 6ea1c23
complete: true
body: read
comments-reviewed: 9
reviews-reviewed: 0
threads-reviewed: 0
threads-unresolved: 0
surfaces-sha256: 05fbac01fe495799c7a64ae9e0f54ebd9430e4852eb2781a2ce08ad487045c3a
diff-files: 3
diff-additions: 20
diff-deletions: 14
checks: latest-wins-green
b0: clear
scope: pass
domain: pass
verdict: READY
organ: check_adjoint_prevalidation.py
organ-command: python scripts/check_adjoint_prevalidation.py --derive-verdict 19675
organ-rc: 0
[/ADJOINT PREFLIGHT]

jsboige · 2026-10-08T01:48:26Z

[ADJOINT PREFLIGHT]
schema: 1
lane: myia-po-2026:CoursIA-3
pr: 19675
head: 6ea1c23
complete: true
body: read
comments-reviewed: 9
reviews-reviewed: 0
threads-reviewed: 0
threads-unresolved: 0
surfaces-sha256: 05fbac01fe495799c7a64ae9e0f54ebd9430e4852eb2781a2ce08ad487045c3a
diff-files: 3
diff-additions: 20
diff-deletions: 14
checks: latest-wins-green
b0: clear
scope: pass
domain: pass
verdict: READY
organ: check_adjoint_prevalidation.py
organ-command: python scripts/check_adjoint_prevalidation.py --derive-verdict 19675
organ-rc: 0
[/ADJOINT PREFLIGHT]

jsboige · 2026-10-08T10:38:01Z

[ADJOINT PREFLIGHT]
schema: 1
lane: myia-ai-01:CoursIA
pr: 19675
head: 6ea1c23
complete: true
body: read
comments-reviewed: 11
reviews-reviewed: 0
threads-reviewed: 0
threads-unresolved: 0
surfaces-sha256: be4211913bc7d4c516b8f9c54a75c87236213664ad75d5c9e7975e201be78422
diff-files: 3
diff-additions: 20
diff-deletions: 14
checks: latest-wins-green
b0: clear
scope: pass
domain: not-applicable
verdict: READY
organ: check_adjoint_prevalidation.py
organ-command: python scripts/check_adjoint_prevalidation.py --derive-verdict 19675
organ-rc: 0
[/ADJOINT PREFLIGHT]

Tier MED/LIGHT, derivation live READY : jambes latest-wins vertes, B.0 clear, aucun thread non resolu ; revues bots lues par le gate.

This was referenced Oct 7, 2026

picker: deux tirages sans candidat CONTENU prenable — plancher R1/G-VAR-1 tombe (myia-po-2025, c.2026-09-17) #16589

Open

[EPIC] Training & Post-Training — trading + sudoku + RL/PPO + GenAI fine-tuning (po-2024 pionnier ⇄ ai-01 approfondit) #1454

Open

jsboige and others added 3 commits October 7, 2026 12:23

Merge branch 'main' into fix/19607-planners-htn-stochasticity

c26f5a7

Merge branch 'main' into fix/19607-planners-htn-stochasticity

27fb0b3

Merge branch 'main' into fix/19607-planners-htn-stochasticity

6ea1c23

myia-ai-01 merged commit 44994e8 into main Oct 8, 2026
97 checks passed

Navigation Menu

Repository navigation

fix(,#19607): Planners-9 HTN -- C.4 stochasticite aries, prose 'MEME plan' etait fausse - #19675

fix(,#19607): Planners-9 HTN -- C.4 stochasticite aries, prose 'MEME plan' etait fausse#19675
myia-ai-01 merged 5 commits into
mainfrom
fix/19607-planners-htn-stochasticity

jsboige commented Oct 7, 2026

Uh oh!

github-actions Bot commented Oct 7, 2026

Uh oh!

github-actions Bot commented Oct 7, 2026

Uh oh!

github-actions Bot commented Oct 7, 2026

Uh oh!

github-actions Bot commented Oct 7, 2026

Uh oh!

github-actions Bot commented Oct 7, 2026

Uh oh!

github-actions Bot commented Oct 7, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 7, 2026

Uh oh!

github-actions Bot commented Oct 7, 2026

Uh oh!

jsboige commented Oct 7, 2026

Uh oh!

jsboige commented Oct 8, 2026

Uh oh!

jsboige commented Oct 8, 2026

Uh oh!

jsboige commented Oct 8, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

jsboige commented Oct 7, 2026

Diagnostic derive

Volet prose -- cellule C# 21 re-alignee

Volet registre -- planners-9-htn.yaml corrige

Pourquoi MD-only (pas de re-exec kernel)

Acceptance (3 cellules-types verifiees)

Lien

Uh oh!

github-actions Bot commented Oct 7, 2026

Uh oh!

github-actions Bot commented Oct 7, 2026

Uh oh!

github-actions Bot commented Oct 7, 2026

Uh oh!

github-actions Bot commented Oct 7, 2026

Uh oh!

github-actions Bot commented Oct 7, 2026

Uh oh!

github-actions Bot commented Oct 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Golden-Set Execution (H.7 P3)

Uh oh!

github-actions Bot commented Oct 7, 2026

Notebook outputs-required (H.4 schema): PASS (every code cell carries an outputs: list)

Uh oh!

github-actions Bot commented Oct 7, 2026

Notebook PR Validation: PASS

Uh oh!

jsboige commented Oct 7, 2026

Uh oh!

jsboige commented Oct 8, 2026

Uh oh!

jsboige commented Oct 8, 2026

Uh oh!

jsboige commented Oct 8, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Volet registre -- `planners-9-htn.yaml` corrige

github-actions Bot commented Oct 7, 2026 •

edited

Loading

Notebook outputs-required (H.4 schema): PASS (every code cell carries an `outputs: list`)