Files

561 lines
12 KiB
YAML

schema_version: 1
study_id: taxonomy-method-two-area-test-2026-11
preregistered_date: '2026-11-02'
random_seed: 20261103
test_areas:
- code: P4
id: serving
label: Deployment & serving
subject_count: 9
subjects:
- P4.1 (serving-architecture)
- P4.2 (compound-ai)
- P4.3 (latency-goodput)
- P4.4 (request-scheduling)
- P4.5 (llm-inference)
- P4.6 (structured-generation)
- P4.7 (test-time-compute)
- P4.8 (retrieval-rag)
- P4.9 (on-device-deployment)
published_0_4_draft_concept_count: 95
unit_counts_by_split:
dev: 142
held_out_test: 116
reserve: 47
unit_counts_by_stratum:
S1: 56
S2: 36
S3: 173
S4: 38
S5: 2
- code: Q4
id: safety-assurance
label: Safety & assurance
subject_count: 4
subjects:
- Q4.1 (hazards-operating-domain)
- Q4.2 (runtime-enforcement)
- Q4.3 (verification-validation)
- Q4.4 (safety-cases)
published_0_4_draft_concept_count: 30
unit_counts_by_split:
dev: 58
held_out_test: 10
reserve: 9
unit_counts_by_stratum:
S1: 55
S2: 8
S3: 6
S4: 5
S5: 3
artifact_hashes:
dev_sources_sha256: 0849d619afb0e745fdde9f5fef715d67cd3d49da31459e2a0b502dafab866df0
held_out_slice_sha256: 221fb94438db2d5e0f78f4dbb73c99c081e73d8788c89cbd12468f6a054607c7
total_units: 382
dev_units: 200
held_out_test_units: 126
reserve_units: 56
enumerated_clusters:
- provenance_family_id: pf_amodei_concrete_safety
area: Q4
stratum: S1
split: dev
unit_count: 1
- provenance_family_id: pf_barroso_wsc
area: P4
stratum: S1
split: dev
unit_count: 1
- provenance_family_id: pf_berkeley_aisafety
area: Q4
stratum: S2
split: held_out_test
unit_count: 1
- provenance_family_id: pf_cmu_llmsys
area: P4
stratum: S2
split: dev
unit_count: 2
- provenance_family_id: pf_cov_courses_2026_AIE
area: P4
stratum: S2
split: dev
unit_count: 4
- provenance_family_id: pf_cov_courses_2026_B294
area: P4
stratum: S2
split: dev
unit_count: 1
- provenance_family_id: pf_cov_courses_2026_BAG
area: Q4
stratum: S2
split: dev
unit_count: 1
- provenance_family_id: pf_cov_courses_2026_C442
area: P4
stratum: S2
split: dev
unit_count: 4
- provenance_family_id: pf_cov_courses_2026_C779
area: P4
stratum: S2
split: dev
unit_count: 1
- provenance_family_id: pf_cov_courses_2026_C868
area: P4
stratum: S2
split: dev
unit_count: 7
- provenance_family_id: pf_cov_courses_2026_DLS
area: P4
stratum: S2
split: dev
unit_count: 1
- provenance_family_id: pf_cov_courses_2026_DMLS
area: P4
stratum: S2
split: dev
unit_count: 1
- provenance_family_id: pf_cov_courses_2026_HAR
area: P4
stratum: S2
split: dev
unit_count: 1
- provenance_family_id: pf_cov_courses_2026_K17
area: P4
stratum: S2
split: dev
unit_count: 2
- provenance_family_id: pf_cov_courses_2026_M594
area: P4
stratum: S2
split: dev
unit_count: 3
- provenance_family_id: pf_cov_courses_2026_MLDP
area: P4
stratum: S2
split: dev
unit_count: 1
- provenance_family_id: pf_cov_courses_2026_MLS26
area: P4
stratum: S2
split: dev
unit_count: 1
- provenance_family_id: pf_cov_courses_2026_NIST
area: Q4
stratum: S2
split: held_out_test
unit_count: 1
- provenance_family_id: pf_cov_courses_2026_OSDI
area: P4
stratum: S2
split: held_out_test
unit_count: 1
- provenance_family_id: pf_cov_courses_2026_RAT
area: P4
stratum: S2
split: dev
unit_count: 1
- provenance_family_id: pf_cov_courses_2026_S329
area: P4
stratum: S2
split: dev
unit_count: 2
- provenance_family_id: pf_cov_courses_2026_S336
area: P4
stratum: S2
split: dev
unit_count: 1
- provenance_family_id: pf_cov_courses_2026_SEI
area: Q4
stratum: S2
split: dev
unit_count: 1
- provenance_family_id: pf_cov_courses_2026_USC
area: Q4
stratum: S2
split: reserve
unit_count: 1
- provenance_family_id: pf_cov_courses_2026_UT378
area: P4
stratum: S2
split: dev
unit_count: 1
- provenance_family_id: pf_cov_mlsys_25_26_mlsys2025
area: P4
stratum: S3
split: dev
unit_count: 16
- provenance_family_id: pf_cov_mlsys_25_26_mlsys2026
area: P4
stratum: S3
split: reserve
unit_count: 45
- provenance_family_id: pf_cov_sysvenues_2025_asplos2025
area: P4
stratum: S3
split: dev
unit_count: 13
- provenance_family_id: pf_cov_sysvenues_2025_eurosys2025
area: P4
stratum: S3
split: dev
unit_count: 8
- provenance_family_id: pf_cov_sysvenues_2025_nsdi2025
area: P4
stratum: S3
split: dev
unit_count: 3
- provenance_family_id: pf_cov_sysvenues_2025_osdi2025
area: P4
stratum: S3
split: dev
unit_count: 2
- provenance_family_id: pf_cov_sysvenues_2025_sosp2025
area: P4
stratum: S3
split: dev
unit_count: 9
- provenance_family_id: pf_cov_sysvenues_2026_heldout_asplos2026
area: P4
stratum: S3
split: held_out_test
unit_count: 25
- provenance_family_id: pf_cov_sysvenues_2026_heldout_eurosys2026
area: P4
stratum: S3
split: held_out_test
unit_count: 20
- provenance_family_id: pf_cov_sysvenues_2026_heldout_nsdi2026
area: P4
stratum: S3
split: held_out_test
unit_count: 12
- provenance_family_id: pf_cov_sysvenues_2026_heldout_osdi2026
area: P4
stratum: S3
split: held_out_test
unit_count: 16
- provenance_family_id: pf_distserve_osdi24
area: P4
stratum: S3
split: dev
unit_count: 1
- provenance_family_id: pf_frontier_scaling_policies
area: Q4
stratum: S4
split: dev
unit_count: 1
- provenance_family_id: pf_huyen_dmls
area: P4
stratum: S1
split: dev
unit_count: 2
- provenance_family_id: pf_iso42001_pas8800
area: Q4
stratum: S5
split: reserve
unit_count: 1
- provenance_family_id: pf_iso_automotive_ul4600
area: Q4
stratum: S5
split: dev
unit_count: 1
- provenance_family_id: pf_k8s_ray_serving
area: P4
stratum: S4
split: dev
unit_count: 1
- provenance_family_id: pf_kalra_statistical_safety
area: Q4
stratum: S3
split: dev
unit_count: 1
- provenance_family_id: pf_katz_verification
area: Q4
stratum: S3
split: dev
unit_count: 1
- provenance_family_id: pf_leveson_stpa
area: Q4
stratum: S1
split: dev
unit_count: 1
- provenance_family_id: pf_mit_han_tinyml
area: P4
stratum: S2
split: dev
unit_count: 1
- provenance_family_id: pf_mlperf_inference
area: P4
stratum: S5
split: dev
unit_count: 2
- provenance_family_id: pf_mlsysbook_v1_v2_vol1/02_system_paradigms
area: P4
stratum: S1
split: dev
unit_count: 2
- provenance_family_id: pf_mlsysbook_v1_v2_vol1/05_nn_computation
area: P4
stratum: S1
split: dev
unit_count: 1
- provenance_family_id: pf_mlsysbook_v1_v2_vol1/06_nn_architectures
area: P4
stratum: S1
split: reserve
unit_count: 1
- provenance_family_id: pf_mlsysbook_v1_v2_vol1/12_benchmarking
area: P4
stratum: S1
split: dev
unit_count: 1
- provenance_family_id: pf_mlsysbook_v1_v2_vol1/13_model_serving
area: P4
stratum: S1
split: held_out_test
unit_count: 5
- provenance_family_id: pf_mlsysbook_v1_v2_vol2/08_fleet_orchestration
area: P4
stratum: S1
split: dev
unit_count: 1
- provenance_family_id: pf_mlsysbook_v1_v2_vol2/10_inference
area: P4
stratum: S1
split: dev
unit_count: 7
- provenance_family_id: pf_mlsysbook_v1_v2_vol2/11_edge_intelligence
area: P4
stratum: S1
split: dev
unit_count: 1
- provenance_family_id: pf_mlsysbook_v1_v2_vol2/14_robust_ai
area: P4
stratum: S1
split: dev
unit_count: 1
- provenance_family_id: pf_mlsysbook_v1_v2_vol2/16_responsible_ai
area: Q4
stratum: S1
split: dev
unit_count: 1
- provenance_family_id: pf_mlsysbook_v3_v4_vol3/02_foundation_model
area: P4
stratum: S1
split: dev
unit_count: 6
- provenance_family_id: pf_mlsysbook_v3_v4_vol3/03_test_time_compute
area: P4
stratum: S1
split: dev
unit_count: 7
- provenance_family_id: pf_mlsysbook_v3_v4_vol3/04_context_engineering
area: P4
stratum: S1
split: dev
unit_count: 1
- provenance_family_id: pf_mlsysbook_v3_v4_vol3/05_kv_cache
area: P4
stratum: S1
split: dev
unit_count: 7
- provenance_family_id: pf_mlsysbook_v3_v4_vol3/06_long_term_memory
area: P4
stratum: S1
split: dev
unit_count: 4
- provenance_family_id: pf_mlsysbook_v3_v4_vol3/10_agent_harness
area: P4
stratum: S1
split: dev
unit_count: 1
- provenance_family_id: pf_mlsysbook_v3_v4_vol3/15_trajectory_curation
area: P4
stratum: S1
split: dev
unit_count: 1
- provenance_family_id: pf_mlsysbook_v3_v4_vol3/17_rlvr
area: P4
stratum: S1
split: dev
unit_count: 1
- provenance_family_id: pf_mlsysbook_v3_v4_vol3/19_agent_economics
area: P4
stratum: S1
split: dev
unit_count: 3
- provenance_family_id: pf_mlsysbook_v3_v4_vol4/01_boundary
area: Q4
stratum: S1
split: dev
unit_count: 4
- provenance_family_id: pf_mlsysbook_v3_v4_vol4/02_body
area: Q4
stratum: S1
split: dev
unit_count: 3
- provenance_family_id: pf_mlsysbook_v3_v4_vol4/03_brain
area: P4
stratum: S1
split: dev
unit_count: 3
- provenance_family_id: pf_mlsysbook_v3_v4_vol4/04_nervous
area: Q4
stratum: S1
split: held_out_test
unit_count: 2
- provenance_family_id: pf_mlsysbook_v3_v4_vol4/05_data
area: Q4
stratum: S1
split: dev
unit_count: 1
- provenance_family_id: pf_mlsysbook_v3_v4_vol4/06_training
area: P4
stratum: S1
split: dev
unit_count: 2
- provenance_family_id: pf_mlsysbook_v3_v4_vol4/07_evaluation
area: Q4
stratum: S1
split: dev
unit_count: 5
- provenance_family_id: pf_mlsysbook_v3_v4_vol4/08_perception
area: Q4
stratum: S1
split: dev
unit_count: 1
- provenance_family_id: pf_mlsysbook_v3_v4_vol4/10_intent
area: Q4
stratum: S1
split: dev
unit_count: 3
- provenance_family_id: pf_mlsysbook_v3_v4_vol4/11_planning
area: Q4
stratum: S1
split: dev
unit_count: 3
- provenance_family_id: pf_mlsysbook_v3_v4_vol4/12_enforcement
area: Q4
stratum: S1
split: dev
unit_count: 7
- provenance_family_id: pf_mlsysbook_v3_v4_vol4/13_placement
area: Q4
stratum: S1
split: dev
unit_count: 1
- provenance_family_id: pf_mlsysbook_v3_v4_vol4/14_intervention
area: Q4
stratum: S1
split: dev
unit_count: 3
- provenance_family_id: pf_mlsysbook_v3_v4_vol4/15_verification
area: Q4
stratum: S1
split: dev
unit_count: 6
- provenance_family_id: pf_mlsysbook_v3_v4_vol4/16_release
area: Q4
stratum: S1
split: reserve
unit_count: 7
- provenance_family_id: pf_mlsysbook_v3_v4_vol4/18_frontier
area: Q4
stratum: S1
split: dev
unit_count: 3
- provenance_family_id: pf_nist_ai_rmf
area: Q4
stratum: S5
split: held_out_test
unit_count: 1
- provenance_family_id: pf_onnx_runtime_docs
area: P4
stratum: S4
split: dev
unit_count: 1
- provenance_family_id: pf_orca_osdi22
area: P4
stratum: S3
split: dev
unit_count: 1
- provenance_family_id: pf_sarathi_osdi24
area: P4
stratum: S3
split: dev
unit_count: 1
- provenance_family_id: pf_sglang_neurips24
area: P4
stratum: S3
split: dev
unit_count: 1
- provenance_family_id: pf_simplex_cbf_control
area: Q4
stratum: S3
split: dev
unit_count: 1
- provenance_family_id: pf_slora_mlsys24
area: P4
stratum: S3
split: held_out_test
unit_count: 1
- provenance_family_id: pf_specdec_icml23
area: P4
stratum: S3
split: dev
unit_count: 1
- provenance_family_id: pf_staffml_heldout_p4
area: P4
stratum: S4
split: held_out_test
unit_count: 36
- provenance_family_id: pf_staffml_heldout_q4
area: Q4
stratum: S4
split: held_out_test
unit_count: 2
- provenance_family_id: pf_staffml_heldout_q4_ext
area: Q4
stratum: S4
split: held_out_test
unit_count: 2
- provenance_family_id: pf_stanford_cs329t
area: Q4
stratum: S2
split: dev
unit_count: 1
- provenance_family_id: pf_stanford_cs336
area: P4
stratum: S2
split: held_out_test
unit_count: 1
- provenance_family_id: pf_ut_cs378
area: P4
stratum: S2
split: reserve
unit_count: 1
- provenance_family_id: pf_vllm_sosp23
area: P4
stratum: S3
split: dev
unit_count: 1
evaluation_metrics_and_thresholds:
held_out_topic_coverage_pct_min: 85.0
seeded_gap_detection_pct_min: 85.0
subject_kappa_min: 0.7
hooper_consistency_pct_min: 40.0
rolling_consistency_pct_min: 57.0
alpha_masi_min: 0.4
max_single_concept_share_pct_max: 15.0
specific_concept_tag_ratio_pct_min: 75.0
vocabulary_size_ratio_bounds:
- 0.7
- 1.35
failure_criterion: The synthesized method does no better than the current method if, on the pre-registered
held-out slice (held_out_test), the rebuilt P4 and Q4 vocabulary fails to improve or match held-out
topic coverage (covered_pct) and multi-rater indexing consistency (hooper_consistency_pct, rolling_consistency_pct,
alpha_masi, subject_kappa) without violating the anti-gaming specificity cap (max_single_concept_share_pct
<= 15.0%) or exceeding 1.35x the published concept count.