KyeongminandClaude Fable 5
4d7c401779
Multi-MDX Regression (IMP-91) / multi-mdx-regression (push) Failing after 9m49s
- tests/integration/scripts/regenerate_snapshots.py: 스냅샷 일괄 재생성 도구 - 통합 스냅샷 9종 갱신, 회귀/유닛 테스트 7건 보강 - ISSUE_DRAFTS_2026-07-02.md: 7/2 전수 검토 이슈 초안 보존 Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
396 lines
8.7 KiB
JSON
396 lines
8.7 KiB
JSON
{
|
|
"_doc": "IMP-91 u7 — F1 V4 ranking observed snapshot (step05_v4_evidence). Pins v4_source (POSIX-normalized), aligned_section_ids, and per-section {section_id, candidate_status, candidates: [{template_id, label, confidence}]}. confidence kept at current 4-decimal rounding. Sections appear in pipeline-emitted order.",
|
|
"01": {
|
|
"v4_source": "tests/matching/v4_full32_result.yaml",
|
|
"aligned_section_ids": [
|
|
"01-intro",
|
|
"01-1",
|
|
"01-2"
|
|
],
|
|
"sections": [
|
|
{
|
|
"section_id": "01-intro",
|
|
"candidate_status": "no_non_reject_v4_candidate",
|
|
"candidates": [
|
|
{
|
|
"template_id": "compensation_complaint_map",
|
|
"label": "reject",
|
|
"confidence": 0.651
|
|
},
|
|
{
|
|
"template_id": "construction_bim_three_usage",
|
|
"label": "reject",
|
|
"confidence": 0.4696
|
|
},
|
|
{
|
|
"template_id": "engn_sw_three_types",
|
|
"label": "reject",
|
|
"confidence": 0.4047
|
|
},
|
|
{
|
|
"template_id": "bim_adoption_central_split",
|
|
"label": "reject",
|
|
"confidence": 0.3994
|
|
},
|
|
{
|
|
"template_id": "bim_dx_comparison_table",
|
|
"label": "reject",
|
|
"confidence": 0.374
|
|
},
|
|
{
|
|
"template_id": "construction_goals_three_circle_intersection",
|
|
"label": "reject",
|
|
"confidence": 0.3418
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"section_id": "01-1",
|
|
"candidate_status": "ok",
|
|
"candidates": [
|
|
{
|
|
"template_id": "construction_bim_three_usage",
|
|
"label": "use_as_is",
|
|
"confidence": 0.9101
|
|
},
|
|
{
|
|
"template_id": "construction_goals_three_circle_intersection",
|
|
"label": "light_edit",
|
|
"confidence": 0.8261
|
|
},
|
|
{
|
|
"template_id": "dx_sw_necessity_three_perspectives",
|
|
"label": "light_edit",
|
|
"confidence": 0.8168
|
|
},
|
|
{
|
|
"template_id": "model_specialized_engn_sw",
|
|
"label": "restructure",
|
|
"confidence": 0.749
|
|
},
|
|
{
|
|
"template_id": "three_parallel_requirements",
|
|
"label": "reject",
|
|
"confidence": 0.7402
|
|
},
|
|
{
|
|
"template_id": "info_management_what_how_when",
|
|
"label": "reject",
|
|
"confidence": 0.7395
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"section_id": "01-2",
|
|
"candidate_status": "ok",
|
|
"candidates": [
|
|
{
|
|
"template_id": "bim_dx_comparison_table",
|
|
"label": "use_as_is",
|
|
"confidence": 0.9459
|
|
},
|
|
{
|
|
"template_id": "process_product_two_way",
|
|
"label": "reject",
|
|
"confidence": 0.8675
|
|
},
|
|
{
|
|
"template_id": "engn_sw_three_types",
|
|
"label": "light_edit",
|
|
"confidence": 0.7571
|
|
},
|
|
{
|
|
"template_id": "app_sw_package_vs_solution",
|
|
"label": "restructure",
|
|
"confidence": 0.6813
|
|
},
|
|
{
|
|
"template_id": "bim_adoption_central_split",
|
|
"label": "restructure",
|
|
"confidence": 0.6455
|
|
},
|
|
{
|
|
"template_id": "industry_characteristics_three_col",
|
|
"label": "reject",
|
|
"confidence": 0.5781
|
|
}
|
|
]
|
|
}
|
|
]
|
|
},
|
|
"02": {
|
|
"v4_source": "tests/matching/v4_full32_result.yaml",
|
|
"aligned_section_ids": [
|
|
"02-1",
|
|
"02-2-sub-1",
|
|
"02-2-sub-2"
|
|
],
|
|
"sections": [
|
|
{
|
|
"section_id": "02-1",
|
|
"candidate_status": "ok",
|
|
"candidates": [
|
|
{
|
|
"template_id": "construction_goals_three_circle_intersection",
|
|
"label": "use_as_is",
|
|
"confidence": 0.914
|
|
},
|
|
{
|
|
"template_id": "three_persona_benefits",
|
|
"label": "reject",
|
|
"confidence": 0.7656
|
|
},
|
|
{
|
|
"template_id": "dx_sw_necessity_three_perspectives",
|
|
"label": "reject",
|
|
"confidence": 0.7512
|
|
},
|
|
{
|
|
"template_id": "model_specialized_engn_sw",
|
|
"label": "reject",
|
|
"confidence": 0.7267
|
|
},
|
|
{
|
|
"template_id": "solution_engn_split_diagram",
|
|
"label": "restructure",
|
|
"confidence": 0.724
|
|
},
|
|
{
|
|
"template_id": "three_parallel_requirements",
|
|
"label": "reject",
|
|
"confidence": 0.6517
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"section_id": "02-2-sub-1",
|
|
"candidate_status": "no_non_reject_v4_candidate",
|
|
"candidates": []
|
|
},
|
|
{
|
|
"section_id": "02-2-sub-2",
|
|
"candidate_status": "ok",
|
|
"candidates": []
|
|
}
|
|
]
|
|
},
|
|
"03": {
|
|
"v4_source": "tests/matching/v4_full32_result.yaml",
|
|
"aligned_section_ids": [
|
|
"03-1",
|
|
"03-2"
|
|
],
|
|
"sections": [
|
|
{
|
|
"section_id": "03-1",
|
|
"candidate_status": "ok",
|
|
"candidates": [
|
|
{
|
|
"template_id": "three_parallel_requirements",
|
|
"label": "use_as_is",
|
|
"confidence": 0.9268
|
|
},
|
|
{
|
|
"template_id": "dx_sw_necessity_three_perspectives",
|
|
"label": "light_edit",
|
|
"confidence": 0.8413
|
|
},
|
|
{
|
|
"template_id": "sw_reality_three_emphasis",
|
|
"label": "reject",
|
|
"confidence": 0.7203
|
|
},
|
|
{
|
|
"template_id": "solution_engn_split_diagram",
|
|
"label": "reject",
|
|
"confidence": 0.6631
|
|
},
|
|
{
|
|
"template_id": "three_persona_benefits",
|
|
"label": "reject",
|
|
"confidence": 0.6548
|
|
},
|
|
{
|
|
"template_id": "industry_current_status_three_col",
|
|
"label": "reject",
|
|
"confidence": 0.6048
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"section_id": "03-2",
|
|
"candidate_status": "ok",
|
|
"candidates": [
|
|
{
|
|
"template_id": "process_product_two_way",
|
|
"label": "use_as_is",
|
|
"confidence": 0.9198
|
|
},
|
|
{
|
|
"template_id": "bim_dx_comparison_table",
|
|
"label": "reject",
|
|
"confidence": 0.8138
|
|
},
|
|
{
|
|
"template_id": "app_sw_package_vs_solution",
|
|
"label": "reject",
|
|
"confidence": 0.6467
|
|
},
|
|
{
|
|
"template_id": "engn_sw_three_types",
|
|
"label": "reject",
|
|
"confidence": 0.6396
|
|
},
|
|
{
|
|
"template_id": "bim_adoption_central_split",
|
|
"label": "restructure",
|
|
"confidence": 0.6278
|
|
},
|
|
{
|
|
"template_id": "industry_current_status_three_col",
|
|
"label": "reject",
|
|
"confidence": 0.5324
|
|
}
|
|
]
|
|
}
|
|
]
|
|
},
|
|
"04": {
|
|
"v4_source": "tests/matching/v4_full32_result.yaml",
|
|
"aligned_section_ids": [
|
|
"04-1",
|
|
"04-2-sub-1",
|
|
"04-2-sub-2"
|
|
],
|
|
"sections": [
|
|
{
|
|
"section_id": "04-1",
|
|
"candidate_status": "no_non_reject_v4_candidate",
|
|
"candidates": [
|
|
{
|
|
"template_id": "pre_construction_model_info_stacked",
|
|
"label": "reject",
|
|
"confidence": 0.7114
|
|
},
|
|
{
|
|
"template_id": "field_effectiveness_five_elements",
|
|
"label": "reject",
|
|
"confidence": 0.6453
|
|
},
|
|
{
|
|
"template_id": "policy_achievement_five_goals",
|
|
"label": "reject",
|
|
"confidence": 0.6412
|
|
},
|
|
{
|
|
"template_id": "compensation_complaint_side_card",
|
|
"label": "reject",
|
|
"confidence": 0.5946
|
|
},
|
|
{
|
|
"template_id": "solution_engn_split_diagram",
|
|
"label": "reject",
|
|
"confidence": 0.4588
|
|
},
|
|
{
|
|
"template_id": "model_specialized_engn_sw",
|
|
"label": "reject",
|
|
"confidence": 0.4502
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"section_id": "04-2-sub-1",
|
|
"candidate_status": "no_non_reject_v4_candidate",
|
|
"candidates": []
|
|
},
|
|
{
|
|
"section_id": "04-2-sub-2",
|
|
"candidate_status": "ok",
|
|
"candidates": []
|
|
}
|
|
]
|
|
},
|
|
"05": {
|
|
"v4_source": "tests/matching/v4_full32_result.yaml",
|
|
"aligned_section_ids": [
|
|
"05-1",
|
|
"05-2"
|
|
],
|
|
"sections": [
|
|
{
|
|
"section_id": "05-1",
|
|
"candidate_status": "ok",
|
|
"candidates": [
|
|
{
|
|
"template_id": "dx_sw_necessity_three_perspectives",
|
|
"label": "light_edit",
|
|
"confidence": 0.77
|
|
},
|
|
{
|
|
"template_id": "model_specialized_engn_sw",
|
|
"label": "restructure",
|
|
"confidence": 0.7333
|
|
},
|
|
{
|
|
"template_id": "sw_reality_three_emphasis",
|
|
"label": "reject",
|
|
"confidence": 0.7265
|
|
},
|
|
{
|
|
"template_id": "three_parallel_requirements",
|
|
"label": "reject",
|
|
"confidence": 0.7198
|
|
},
|
|
{
|
|
"template_id": "solution_engn_split_diagram",
|
|
"label": "restructure",
|
|
"confidence": 0.7011
|
|
},
|
|
{
|
|
"template_id": "industry_characteristics_three_col",
|
|
"label": "reject",
|
|
"confidence": 0.6693
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"section_id": "05-2",
|
|
"candidate_status": "no_non_reject_v4_candidate",
|
|
"candidates": [
|
|
{
|
|
"template_id": "compensation_complaint_map",
|
|
"label": "reject",
|
|
"confidence": 0.5526
|
|
},
|
|
{
|
|
"template_id": "design_method_distortion_three_col",
|
|
"label": "reject",
|
|
"confidence": 0.3993
|
|
},
|
|
{
|
|
"template_id": "engn_sw_three_types",
|
|
"label": "reject",
|
|
"confidence": 0.3487
|
|
},
|
|
{
|
|
"template_id": "bigroom_system_components",
|
|
"label": "reject",
|
|
"confidence": 0.3364
|
|
},
|
|
{
|
|
"template_id": "sw_dependency_four_problems",
|
|
"label": "reject",
|
|
"confidence": 0.2968
|
|
},
|
|
{
|
|
"template_id": "app_sw_package_vs_solution",
|
|
"label": "reject",
|
|
"confidence": 0.2555
|
|
}
|
|
]
|
|
}
|
|
]
|
|
}
|
|
}
|