gp3ml 0.3.2 — release candidate (not published)

October 2026 documentation and reproducibility

• Adds a non-exported grouped-versus-row holdout leakage audit, synthetic paired-ledger visualizations, and deterministic tests. No exported ML API or transportability guarantee is changed.
• Release gate: complete exact-source R CMD check –as-cran, review installed source tarball and cross-platform tests, audit publication metadata, then publish an immutable GitHub tag and submit to CRAN with maintainer confirmation.
• Current CRAN baseline at candidate preparation: 0.3.0. The candidate has not been submitted to CRAN or published as a GitHub Release.

gp3ml 0.3.1

Documentation and release maintenance

• Added the September 2026 ecosystem methods update and registered it in the pkgdown article navigation and article index.
• Refreshed the generated public documentation and website metadata since the 0.3.0 source release.
• Preserved the established scientific implementation, public API contracts, governance semantics, and backward compatibility; this patch release does not add or alter R modelling logic.

gp3ml 0.3.0

CRAN availability and metadata

• Added the canonical CRAN package page, CRAN installation instructions, package-status badge, and CRAN package DOI following publication of gp3ml 0.1.0 on CRAN.

Next-generation governed research-ML infrastructure

• Added explicit decision-rule contracts, threshold evaluation, deterministic threshold selection, and abstention auditing without hidden threshold defaults.

• Added target-aware split-conformal calibration for regression intervals and binary prediction sets, with explicit calibration-unit and generalization-target semantics.

• Added predictor-distribution and missingness-shift audits that remain distinct from prevalence, calibration, and performance drift.

• Added frozen analysis-plan contracts with SHA-256 locking and post hoc deviation auditing.

• Added portable model artifacts with optional bundle support, serialization validation, and optional fresh-process checks through callr.

• Added robustness and stability diagnostics across seeds, features, thresholds, and missingness scenarios.

• Added environment capture/comparison, minimal RO-Crate-oriented research-object export, SHA-256 release manifests, and a manual GitHub release-attestation workflow.

• Added gp3ml-native, NIST AI RMF 1.0, ISO/IEC 23894-oriented, and ISO/IEC 42001-oriented governance evidence profiles. Standards-oriented outputs explicitly do not claim certification or conformity.

• Added six new development articles covering decision governance, conformal prediction, dataset shift and robustness, frozen analysis plans, portable research artifacts, and governance standards crosswalks.

• Added an explicit API-stability registry, public-object schema audit, and contract validation for the established 0.2.0 surface.

• Added lightweight cross-package handoff contracts for gp3tools, gpbiometrics, gp3sequences, study-design data, and custom prepared data without duplicating upstream preprocessing.

• Added deterministic documentation-output controls plus audits for volatile timestamps, temporary paths, and runtime addresses.

• Added engine capability auditing and a Windows/macOS/Linux optional-engine CI matrix.

• Added a realistic synthetic three-package research handoff and governed participant-generalization validation workflow.

• Added five package-maturity vignettes covering API contracts, interoperability, reproducibility, portability, and integrated workflow validation.

gp3ml 0.2.0

Grouped evaluation and governed tuning

• Added fold-local evaluation across materialized gazepoint_group_folds, with retained predictions, metrics, exclusions, leakage audits, warnings, and failures.
• Added explicit deterministic tuning grids, candidate comparison, and human-reviewed selection requiring a predeclared metric and direction.
• Added nested grouped resampling with inner tuning isolated inside every outer analysis partition.

Target-aligned uncertainty

• Added observation, participant-cluster, stimulus-cluster, and participant-and-stimulus bootstrap uncertainty with the resampling unit recorded explicitly.
• Added fold- and repeat-distribution summaries without relabelling row-level results as cluster-level estimates.

External validation and reporting

• Added independent-dataset declarations, schema checks, predictor availability, prevalence shift, calibration drift, participant/stimulus coverage, and transportability limitations.
• Added explicit not_externally_validated status when no independent dataset is available.
• Added release-ready model cards recording selection procedures, uncertainty units, transportability status, limitations, and human oversight.

Vignettes and validation

• Added nine deterministic synthetic vignettes covering recording quality, assigned conditions, observed outcomes, participant and stimulus generalization, simultaneous participant/stimulus generalization, contaminated manifests, nested resampling, and external validation.
• Added focused tests, an end-to-end smoke test, a full validation driver, and a read-only 0.2.0 release audit.

gp3ml 0.1.0

First formal release

• Established gp3ml as a governance-first package for leakage-resistant predictive modelling and validation using Gazepoint-derived research data.
• Restricted supported tasks to explicitly observed, non-sensitive outcomes and clearly declared scientific purposes.
• Added explicit prohibited-use documentation covering identification, authentication, diagnostic, protected-attribute, emotion, stress, personality, deception, cognition, comprehension, intent, and other mental-state inference.

Governance, provenance, and leakage protection

• Added task declaration, use-case assertion, variable-role validation, and machine-readable prohibited-use helpers.
• Added feature-provenance manifests covering predictor origins, transformations, availability stages, roles, and preprocessing scope.
• Added structured leakage audits for row, participant, participant-trial, stimulus, identifier, target-derived, and post-outcome risks.

Group-aware validation

• Added deterministic group-aware holdout splitting and repeated grouped resampling for new trials among known participants, new participants, new stimuli, and simultaneous new-participant and new-stimulus generalization.
• Materialized analysis, assessment, and explicitly excluded partitions with complete source-row accounting and embedded leakage audits.
• Added fold-balance, coverage, exclusion, and outcome-representation diagnostics with structured pass, review, and fail findings.

Governed modelling core

• Added fold-local preprocessing objects with separate fitting and baking interfaces.
• Added governed model engines, explicit black-box integration, classification and regression metrics, calibration assessment, and explicitly labelled bootstrap metric uncertainty.
• Added external-validation evaluation and reporting without treating an internal holdout as external validation.
• Added model cards and reproducibility reports that record task purpose, governance decisions, model settings, performance, calibration, uncertainty, limitations, and reproducibility information.
• Model selection remains explicit and reviewable; the package does not perform autonomous black-box winner selection.

Documentation and quality assurance

• Added machine-readable CSV, Markdown, and JSON reporting interfaces where supported by the relevant object.
• Added deterministic synthetic tests across supported generalization targets, governance failures, leakage cases, model engines, metrics, calibration, reporting, and serialization.
• Added complete pkgdown reference organization for the first formal release.