Benchmark Rerun Kits¶

A benchmark rerun kit connects three independently reviewable things: a governed Core asset, a public Runtime entrypoint, and a recorded result bundle. A manifest without execution is only an inspectable corpus; an execution without a manifest is an unattributed run; a run bundle without comparison policy cannot support a parity claim.

flowchart LR
    manifest["Core package manifest"] --> input["identified benchmark inputs"]
    input --> entry["Runtime entrypoint"]
    entry --> run["run bundle + artifact inventory"]
    run --> compare["declared comparison policy"]
    compare --> posture{"claim posture"}
    posture -->|accepted| bounded["bounded family evidence"]
    posture -->|failed| refusal["failure or refusal record"]

Family Rerun Routes¶

All entrypoints below are importable from bijux_proteomics_runtime.workflows. The primary lane reopens the flagship package; the companion lane applies transfer or stress pressure from a distinct governed package.

Family	Primary package	Primary entrypoint	Primary mode	Companion package	Companion entrypoint	Companion mode
`dda`	`dda_reviewable_run`	`paths.run_reviewable_import_path`	`import_only`	`dda_cross_engine_review_package`	`benchmark_runs.run_benchmark_dda_generalization_import_path`	`import_only`
`dia`	`dia_library_review_package`	`benchmark_runs.run_benchmark_dia_review_path`	`raw_executable`	`dia_matrix_shift_review_package`	`benchmark_runs.run_benchmark_dia_generalization_review_path`	`raw_executable`
`lfq`	`lfq_cohort_review_package`	`benchmark_runs.run_benchmark_lfq_review_path`	`raw_executable`	`lfq_sparse_contrast_review_package`	`benchmark_runs.run_benchmark_lfq_generalization_review_path`	`raw_executable`
`multiplex`	`multiplex_tmtpro_review_package`	`benchmark_runs.run_benchmark_multiplex_review_path`	`raw_executable`	`multiplex_channel_stress_review_package`	`benchmark_runs.run_benchmark_multiplex_generalization_review_path`	`raw_executable`
`ptm`	`ptm_localization_review_package`	`benchmark_runs.run_benchmark_ptm_review_path`	`raw_executable`	`ptm_ambiguity_stress_review_package`	`benchmark_runs.run_benchmark_ptm_generalization_review_path`	`raw_executable`
`targeted`	`targeted_transition_review_package`	`benchmark_runs.run_benchmark_targeted_review_path`	`raw_executable`	`targeted_carryover_review_package`	`benchmark_runs.run_benchmark_targeted_generalization_review_path`	`raw_executable`

If a future family has no governed companion lane, its entry must say not published for this family; a primary rerun must never imply generalization evidence.

DDA is deliberately different from the other lanes. Its primary route imports a checked MaxQuant result rather than running an in-repository search engine. Its companion imports a distinct Comet/Sage comparison package, adding cross-engine pressure without turning imported execution into a native-search claim.

Open A Kit¶

From a clean checkout and installed workspace environment:

Open the package package_manifest.json and identify its scientific family, source locator, expected inventory, and declared run mode.
Verify the files named by artifact_inventory.json and their checksums.
Call the family's primary entrypoint and write its result below artifacts/bijux-proteomics-runtime/.
Preserve the resolved configuration, input identities, provider, terminal state, diagnostics, and output hashes.
Call the companion entrypoint independently.
Compare the results only under the rule in the Benchmark Comparability Matrix.
Inspect the family refusal before writing a stronger claim.

Do not overwrite tracked fixtures beneath packages/bijux-proteomics-runtime/tests/fixtures/flagship_runs/. They are governed test evidence; a new run remains under artifacts/ until promotion is explicitly reviewed.

Family Evidence¶

`dda`¶

requested public language: outsider_auditable_bounded
black-box allowed language: review_grade_bounded
primary package root: packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dda_reviewable_run
companion package root: packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dda_cross_engine_review_package
primary runtime entrypoint: bijux_proteomics_runtime.workflows.paths.run_reviewable_import_path
primary run mode: import_only
companion runtime entrypoint: bijux_proteomics_runtime.workflows.benchmark_runs.run_benchmark_dda_generalization_import_path
companion run mode: import_only

Opening order:

packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dda_reviewable_run/package_manifest.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dda_reviewable_run/artifact_inventory.json
packages/bijux-proteomics-core/tests/fixtures/search_adapter_corpora/maxquant/maxquant_pipeline_export.tsv
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dda_cross_engine_review_package/package_manifest.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dda_cross_engine_review_package/artifact_inventory.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dda_cross_engine_review_package/primary/comet_pipeline_export.tsv
artifacts/intelligence/independent-reruns/dda_independent_rerun_dossier.json
artifacts/intelligence/external-review-kits/dda_external_review_kit.json

Validating tests:

packages/bijux-proteomics-runtime/tests/workflows/test_runtime_external_pack_surface.py
packages/bijux-proteomics-runtime/tests/workflows/test_benchmark_runtime_surface.py
independent rerun dossier: artifacts/intelligence/independent-reruns/dda_independent_rerun_dossier.json
external review kit: artifacts/intelligence/external-review-kits/dda_external_review_kit.json

Remaining limits:

no in-repo live-engine rerun parity
one-run package cannot authorize broad production-cohort DDA claims
no live-engine rerun parity
generalization remains bounded to two small exported-result packages
The dossier still describes repository-owned rerun lanes, not an untracked third-party reproduction outside the repository boundary.
The strongest shipped rerun lane is still import-backed rather than a raw external-engine execution owned by this repository.

Open the primary package first, then the companion package, then the rerun dossier or external review kit if one exists. The kit is meant to survive without maintainer narration.

`dia`¶

requested public language: outsider_auditable_bounded
black-box allowed language: outsider_auditable_bounded
primary package root: packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dia_library_review_package
companion package root: packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dia_matrix_shift_review_package
primary runtime entrypoint: bijux_proteomics_runtime.workflows.benchmark_runs.run_benchmark_dia_review_path
primary run mode: raw_executable
companion runtime entrypoint: bijux_proteomics_runtime.workflows.benchmark_runs.run_benchmark_dia_generalization_review_path
companion run mode: raw_executable

Opening order:

packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dia_library_review_package/package_manifest.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dia_library_review_package/artifact_inventory.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dia_library_review_package/primary/spectronaut_report.tsv
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dia_matrix_shift_review_package/package_manifest.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dia_matrix_shift_review_package/artifact_inventory.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dia_matrix_shift_review_package/primary/diann_report.tsv
artifacts/intelligence/independent-reruns/dia_independent_rerun_dossier.json
artifacts/intelligence/external-review-kits/dia_external_review_kit.json

Validating tests:

packages/bijux-proteomics-runtime/tests/workflows/test_runtime_external_pack_surface.py
packages/bijux-proteomics-runtime/tests/workflows/test_benchmark_runtime_surface.py
independent rerun dossier: artifacts/intelligence/independent-reruns/dia_independent_rerun_dossier.json
external review kit: artifacts/intelligence/external-review-kits/dia_external_review_kit.json

Remaining limits:

no chromatogram-level vendor parity
library incompleteness and absent-peptide consequences still block broader biological confidence
protein-evidence transfer remains weaker than precursor-level review transfer
library-conditioned authority still caps the family posture
The dossier still describes repository-owned rerun lanes, not an untracked third-party reproduction outside the repository boundary.
The strongest shipped rerun lane is raw-executable inside the repository, but vendor-parity and broader ecosystem replay are still separate questions.

Open the primary package first, then the companion package, then the rerun dossier or external review kit if one exists. The kit is meant to survive without maintainer narration.

`lfq`¶

requested public language: outsider_auditable_bounded
black-box allowed language: outsider_auditable_bounded
primary package root: packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/lfq_cohort_review_package
companion package root: packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/lfq_sparse_contrast_review_package
primary runtime entrypoint: bijux_proteomics_runtime.workflows.benchmark_runs.run_benchmark_lfq_review_path
primary run mode: raw_executable
companion runtime entrypoint: bijux_proteomics_runtime.workflows.benchmark_runs.run_benchmark_lfq_generalization_review_path
companion run mode: raw_executable

Opening order:

packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/lfq_cohort_review_package/package_manifest.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/lfq_cohort_review_package/artifact_inventory.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/lfq_cohort_review_package/evidence/study_scale_ms1_features.tsv
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/lfq_sparse_contrast_review_package/package_manifest.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/lfq_sparse_contrast_review_package/artifact_inventory.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/lfq_sparse_contrast_review_package/evidence/edge_case_ms1_features.tsv
artifacts/intelligence/independent-reruns/lfq_independent_rerun_dossier.json
artifacts/intelligence/external-review-kits/lfq_external_review_kit.json

Validating tests:

packages/bijux-proteomics-runtime/tests/workflows/test_flagship_run_bundle_surface.py
packages/bijux-proteomics-runtime/tests/workflows/test_benchmark_runtime_surface.py
independent rerun dossier: artifacts/intelligence/independent-reruns/lfq_independent_rerun_dossier.json
external review kit: artifacts/intelligence/external-review-kits/lfq_external_review_kit.json

Remaining limits:

no stronger public truth package for accuracy beyond repeatability
generalization beyond the current cohort package remains explicitly bounded
effect-direction confidence weakens under sparser contrast
family authority remains bounded rather than decision-grade
The dossier still describes repository-owned rerun lanes, not an untracked third-party reproduction outside the repository boundary.
The strongest shipped rerun lane is raw-executable inside the repository, but vendor-parity and broader ecosystem replay are still separate questions.

Open the primary package first, then the companion package, then the rerun dossier or external review kit if one exists. The kit is meant to survive without maintainer narration.

`multiplex`¶

requested public language: internal_support_only
black-box allowed language: internal_support_only
primary package root: packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/multiplex_tmtpro_review_package
companion package root: packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/multiplex_channel_stress_review_package
primary runtime entrypoint: bijux_proteomics_runtime.workflows.benchmark_runs.run_benchmark_multiplex_review_path
primary run mode: raw_executable
companion runtime entrypoint: bijux_proteomics_runtime.workflows.benchmark_runs.run_benchmark_multiplex_generalization_review_path
companion run mode: raw_executable

Opening order:

packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/multiplex_tmtpro_review_package/package_manifest.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/multiplex_tmtpro_review_package/artifact_inventory.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/multiplex_tmtpro_review_package/evidence/multiplex_ms1_features.tsv
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/multiplex_channel_stress_review_package/package_manifest.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/multiplex_channel_stress_review_package/artifact_inventory.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/multiplex_channel_stress_review_package/evidence/multiplex_channel_stress_ms1_features.tsv

Validating tests:

packages/bijux-proteomics-runtime/tests/workflows/test_flagship_run_bundle_surface.py
packages/bijux-proteomics-runtime/tests/workflows/test_benchmark_runtime_surface.py
independent rerun dossier: not published for this family
external review kit: not published for this family

Remaining limits:

no multiplex lab packet or outsider decision brief family
multiplex authority is intentionally kept out of the outsider-facing flagship set
multiplex still lacks outsider review and lab consequence posture
public release language remains internal-support only even with a second package
Multiplex remains internal-support only, so the rerun kit is reviewable but not a route to outsider-auditable language.

Open the primary package first, then the companion package, then the rerun dossier or external review kit if one exists. The kit is meant to survive without maintainer narration.

`ptm`¶

requested public language: outsider_auditable_bounded
black-box allowed language: outsider_auditable_bounded
primary package root: packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/ptm_localization_review_package
companion package root: packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/ptm_ambiguity_stress_review_package
primary runtime entrypoint: bijux_proteomics_runtime.workflows.benchmark_runs.run_benchmark_ptm_review_path
primary run mode: raw_executable
companion runtime entrypoint: bijux_proteomics_runtime.workflows.benchmark_runs.run_benchmark_ptm_generalization_review_path
companion run mode: raw_executable

Opening order:

packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/ptm_localization_review_package/package_manifest.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/ptm_localization_review_package/artifact_inventory.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/ptm_localization_review_package/evidence/localization_results.tsv
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/ptm_ambiguity_stress_review_package/package_manifest.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/ptm_ambiguity_stress_review_package/artifact_inventory.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/ptm_ambiguity_stress_review_package/evidence/localization_results.tsv
artifacts/intelligence/independent-reruns/ptm_independent_rerun_dossier.json
artifacts/intelligence/external-review-kits/ptm_external_review_kit.json

Validating tests:

packages/bijux-proteomics-runtime/tests/workflows/test_flagship_run_bundle_surface.py
packages/bijux-proteomics-runtime/tests/workflows/test_benchmark_runtime_surface.py
independent rerun dossier: artifacts/intelligence/independent-reruns/ptm_independent_rerun_dossier.json
external review kit: artifacts/intelligence/external-review-kits/ptm_external_review_kit.json

Remaining limits:

occupancy and regulatory interpretation still remain narrower than localization evidence
PTM follow-up remains exploratory and bounded by ambiguity-aware consequence planning
targetability weakens materially under ambiguity stress
family authority remains bounded rather than decision-grade
The dossier still describes repository-owned rerun lanes, not an untracked third-party reproduction outside the repository boundary.
The strongest shipped rerun lane is raw-executable inside the repository, but vendor-parity and broader ecosystem replay are still separate questions.

Open the primary package first, then the companion package, then the rerun dossier or external review kit if one exists. The kit is meant to survive without maintainer narration.

`targeted`¶

requested public language: outsider_auditable_bounded
black-box allowed language: outsider_auditable_bounded
primary package root: packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/targeted_transition_review_package
companion package root: packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/targeted_carryover_review_package
primary runtime entrypoint: bijux_proteomics_runtime.workflows.benchmark_runs.run_benchmark_targeted_review_path
primary run mode: raw_executable
companion runtime entrypoint: bijux_proteomics_runtime.workflows.benchmark_runs.run_benchmark_targeted_generalization_review_path
companion run mode: raw_executable

Opening order:

packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/targeted_transition_review_package/package_manifest.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/targeted_transition_review_package/artifact_inventory.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/targeted_transition_review_package/evidence/targeted_benchmark_qc.tsv
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/targeted_carryover_review_package/package_manifest.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/targeted_carryover_review_package/artifact_inventory.json
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/targeted_carryover_review_package/evidence/targeted_benchmark_qc.tsv
artifacts/intelligence/independent-reruns/targeted_independent_rerun_dossier.json
artifacts/intelligence/external-review-kits/targeted_external_review_kit.json

Validating tests:

packages/bijux-proteomics-runtime/tests/workflows/test_flagship_run_bundle_surface.py
packages/bijux-proteomics-runtime/tests/workflows/test_benchmark_runtime_surface.py
independent rerun dossier: artifacts/intelligence/independent-reruns/targeted_independent_rerun_dossier.json
external review kit: artifacts/intelligence/external-review-kits/targeted_external_review_kit.json

Remaining limits:

vendor-parity and calibration-clean authority are still outside the current proof boundary
targeted follow-up remains exploratory and cannot authorize calibration-perfect biological certainty
stronger carryover pressure weakens promotion confidence
family authority remains bounded by calibration and vendor-parity limits
The dossier still describes repository-owned rerun lanes, not an untracked third-party reproduction outside the repository boundary.
The strongest shipped rerun lane is raw-executable inside the repository, but vendor-parity and broader ecosystem replay are still separate questions.

Open the primary package first, then the companion package, then the rerun dossier or external review kit if one exists. The kit is meant to survive without maintainer narration.

Read The Result Bundle¶

Evidence	Question it answers	Question it cannot answer alone
benchmark manifest	which corpus and family contract were requested?	did execution complete?
runtime state history	which states and refusals occurred?	are the scientific outputs acceptable?
artifact inventory and hashes	which outputs were produced without substitution?	are two outputs scientifically equivalent?
environment record	which provider and dependencies shaped execution?	will another environment behave identically?
comparison report	which declared fields remained stable?	does the result generalize outside the corpus?
Core acceptance result	did the output meet family-specific bars?	is the biological interpretation grounded?

The Black-Box Benchmark Dashboard summarizes installed-entrypoint checks. The Flagship Run Registry binds published run identities to artifacts. Neither replaces the underlying bundle.

Requested And Allowed Language¶

Family	Primary mode	Requested language	Black-box allowed language	Limits that remain visible
`dda`	`import_only`	`outsider_auditable_bounded`	`review_grade_bounded`	no in-repo live-engine rerun parity; one-run package cannot authorize broad production-cohort DDA claims
`dia`	`raw_executable`	`outsider_auditable_bounded`	`outsider_auditable_bounded`	no chromatogram-level vendor parity; library incompleteness and absent-peptide consequences still block broader biological confidence
`lfq`	`raw_executable`	`outsider_auditable_bounded`	`outsider_auditable_bounded`	no stronger public truth package for accuracy beyond repeatability; generalization beyond the current cohort package remains explicitly bounded
`multiplex`	`raw_executable`	`internal_support_only`	`internal_support_only`	no multiplex lab packet or outsider decision brief family; multiplex authority is intentionally kept out of the outsider-facing flagship set
`ptm`	`raw_executable`	`outsider_auditable_bounded`	`outsider_auditable_bounded`	occupancy and regulatory interpretation still remain narrower than localization evidence; PTM follow-up remains exploratory and bounded by ambiguity-aware consequence planning
`targeted`	`raw_executable`	`outsider_auditable_bounded`	`outsider_auditable_bounded`	vendor-parity and calibration-clean authority are still outside the current proof boundary; targeted follow-up remains exploratory and cannot authorize calibration-perfect biological certainty

Runtime completion proves operational execution under recorded inputs and environment. It does not prove source authenticity, scientific acceptance, grounded biological truth, recommendation authority, or laboratory value.

Continue The Audit¶

Runtime Execution Boundary gives the manifest, entrypoint, fixture, and refusal for every primary lane.
Runtime Replay Challenges applies state, environment, and artifact perturbations.
Raw Versus Import Execution distinguishes native computation from custody of external results.
Runtime Rerun Refusals states the evidence needed before each claim can widen.
Benchmark Assets covers provenance, redistribution, freshness, and incompleteness.

Benchmark Rerun Kits¶

Family Rerun Routes¶

Open A Kit¶

Family Evidence¶

dda¶

dia¶

lfq¶

multiplex¶

ptm¶

targeted¶

Read The Result Bundle¶

Requested And Allowed Language¶

Continue The Audit¶

`dda`¶

`dia`¶

`lfq`¶

`multiplex`¶

`ptm`¶

`targeted`¶