1 datasets.bundles.posthoc_context

datasets.bundles.posthoc_context

Bind campaign context to post-hoc rollout evidence without reopening labels.

The rollout extractor owns candidate and state observations. This module owns the narrower campaign join: it authenticates the input catalog and frozen scene assignment, attaches only root rows referenced by admitted chains, and records the disposition and timing of every catalog member. Validation population metadata may be retained as dispositions while a TRAIN-only analysis remains free of validation observations.

1.1 Functions

Name	Description
bind_archived_q1_context	Attach sealed Q1 state metrics only after exact native-label admission.
bind_posthoc_context	Bind a frozen campaign catalog, population assignment, roots and costs.
freeze_nomination	Attach at most two authored TRAIN nominations and publish new evidence.

1.1.1 bind_archived_q1_context

datasets.bundles.posthoc_context.bind_archived_q1_context(
    evidence,
    archive_path,
    expected_manifest_sha256,
    *,
    study_prefix='.data/architecture_research/20261006/first-model-study',
)

Attach sealed Q1 state metrics only after exact native-label admission.

The archived owner validates the study and computes its existing post-hoc frame. This adapter admits only archived R1 stores present in the evidence, joins by manifest/rollout/state identity, and requires exact shell-index and raw-label agreement before carrying model metrics across the boundary. Missing R0 stores remain absent rather than receiving inferred predictions.

1.1.2 bind_posthoc_context

datasets.bundles.posthoc_context.bind_posthoc_context(
    evidence,
    catalog_path,
    execution,
    *,
    include_roots=True,
)

Bind a frozen campaign catalog, population assignment, roots and costs.

Only root samples referenced by admitted chains are read. Each immutable root manifest is hashed once and summarized once for the union of referenced source indices. Catalog members that are not part of the admitted TRAIN product remain explicit dispositions rather than becoming observations.

1.1.3 freeze_nomination

datasets.bundles.posthoc_context.freeze_nomination(evidence, changes, output)

Attach at most two authored TRAIN nominations and publish new evidence.

This operation records qualitative proposals only. It rejects validation observations and effect/estimate fields, and never selects a policy.

# datasets.bundles.posthoc_context { #aria_nbv.datasets.bundles.posthoc_context } `datasets.bundles.posthoc_context` Bind campaign context to post-hoc rollout evidence without reopening labels. The rollout extractor owns candidate and state observations. This module owns the narrower campaign join: it authenticates the input catalog and frozen scene assignment, attaches only root rows referenced by admitted chains, and records the disposition and timing of every catalog member. Validation population metadata may be retained as dispositions while a TRAIN-only analysis remains free of validation observations. ## Functions | Name | Description | | --- | --- | | [bind_archived_q1_context](#aria_nbv.datasets.bundles.posthoc_context.bind_archived_q1_context) | Attach sealed Q1 state metrics only after exact native-label admission. | | [bind_posthoc_context](#aria_nbv.datasets.bundles.posthoc_context.bind_posthoc_context) | Bind a frozen campaign catalog, population assignment, roots and costs. | | [freeze_nomination](#aria_nbv.datasets.bundles.posthoc_context.freeze_nomination) | Attach at most two authored TRAIN nominations and publish new evidence. | ### bind_archived_q1_context { #aria_nbv.datasets.bundles.posthoc_context.bind_archived_q1_context } ```python datasets.bundles.posthoc_context.bind_archived_q1_context( evidence, archive_path, expected_manifest_sha256, *, study_prefix='.data/architecture_research/20261006/first-model-study', ) ``` Attach sealed Q1 state metrics only after exact native-label admission. The archived owner validates the study and computes its existing post-hoc frame. This adapter admits only archived R1 stores present in the evidence, joins by manifest/rollout/state identity, and requires exact shell-index and raw-label agreement before carrying model metrics across the boundary. Missing R0 stores remain absent rather than receiving inferred predictions. ### bind_posthoc_context { #aria_nbv.datasets.bundles.posthoc_context.bind_posthoc_context } ```python datasets.bundles.posthoc_context.bind_posthoc_context( evidence, catalog_path, execution, *, include_roots=True, ) ``` Bind a frozen campaign catalog, population assignment, roots and costs. Only root samples referenced by admitted chains are read. Each immutable root manifest is hashed once and summarized once for the union of referenced source indices. Catalog members that are not part of the admitted TRAIN product remain explicit dispositions rather than becoming observations. ### freeze_nomination { #aria_nbv.datasets.bundles.posthoc_context.freeze_nomination } ```python datasets.bundles.posthoc_context.freeze_nomination(evidence, changes, output) ``` Attach at most two authored TRAIN nominations and publish new evidence. This operation records qualitative proposals only. It rejects validation observations and effect/estimate fields, and never selects a policy.