用途:对当前 package 中所有 figure 的用途、路径和引用方式做一份统一说明。 原则:本 atlas 既回答“哪些图被用了”,也回答“每张图在当前论证中负责什么”。
当前 Phase 4 + next-stage + next-layer 的叙事既依赖旧的 anchor figures,也依赖这次新生成的状态/bridge/gate 图;同时,meeting 里已经明确要求主图、补充图、历史图、archive support figure 分层管理。因此,如果只给 PDF,而不给单独 atlas,后来的人很容易知道“图存在”,但不知道:
6314629F1 | results/phase4plus_figures/phase4plus_fig1_delivery_status_matrix.png | Top-level state ladder from frozen Phase 4 to evidence-gated delivery.F2 | results/phase4plus_figures/phase4plus_fig2_cross_country_bridge.png | Accuracy and delta profile for UK, USA, and MTUS wave-1 bridge rows.F3 | results/phase4plus_figures/phase4plus_fig3_mtus_wave1_country_delta.png | Five-country wave-1 delta profile for current MTUS bridge evidence.F4 | results/phase4plus_figures/phase4plus_fig4_groupb_axes.png | Employment, gender, income, and age status with structural-boundary annotation.F5 | results/phase4plus_figures/phase4plus_fig5_extension_support.png | Trigger difficulty, contextual/hazard lift, and expanding-window negativity.F6 | results/phase4plus_figures/phase4plus_fig6_evidence_gate.png | Default-delivery pass/hold structure and stronger-claim trigger rule.ANCHOR-6 | results/anchor_figures/figure6_transition_analysis.png | Primary mechanism-first anchor for stay versus transition asymmetry.ANCHOR-8 | results/anchor_figures/figure8_input_info_effect.png | Anchor figure for the claim that more history or channels do not materially improve prediction.ANCHOR-10 | results/anchor_figures/figure10_dimension_importance.png | Primary social-stratification figure, but only under explicit caption boundaries.FIGURE1 | results/project_figures/figure1_a_class_performance.png | Historical stage overview for cross-channel asymmetry. Keep as background, not current mainline.FIGURE2 | results/project_figures/figure2_b_class_heatmap.png | Grouped heatmap context for older subgroup evidence.FIGURE3 | results/project_figures/figure3_e2b_waterfall.png | Historical channel-to-channel asymmetry support figure.FIGURE4 | results/project_figures/figure4_e1_model_comparison.png | Historical accuracy-first comparison. Preserve as archive evidence only.FIGURE5 | results/project_figures/figure5_bootstrap_ci.png | Bootstrap uncertainty support for between-group spread estimates.FIGURE7 | results/project_figures/figure7_imputation_robustness.png | Supplementary robustness figure for missingness handling.FIGURE9 | results/project_figures/figure9_unified_model_comparison.png | Historical unified comparison figure retained as background.FIGURE11 | results/project_figures/figure11_deep_transition.png | Current mainline support figure for transition difficulty across model families.FIGURE12 | results/project_figures/figure12_order_vs_bag.png | Current mechanism-support figure showing sequence order dominance.FIGURE13 | results/project_figures/figure13_role_timelines.png | Interpretive timeline figure for role-pattern communication.FIGURE14 | results/project_figures/figure14_role_comparison.png | Interpretive comparison figure for role-pattern communication.SUPP1 | results/project_figures/supplementary_fig1_deep_models_by_income.png | Supplementary figure for high-income complexity and deep-model lift.SUPP2 | results/project_figures/supplementary_fig2_multitask_comparison.png | Supplementary architecture-comparison figure.SUPP3 | results/project_figures/supplementary_fig3_error_rate_by_group.png | Supplementary group-level error distribution figure.P2-F8 | results/archive_support_figures/phase2_fig8_us_sample_size_delta_shift.png | Archive support figure showing why quick negatives should not be over-read without pooled support.P2-F10 | results/archive_support_figures/phase2_fig10_us_pooled_support_vs_delta.png | Archive support figure for pooled support-width and activity/location delta interpretation.P3-F1 | results/archive_support_figures/phase3_fig1_a1_delta_by_country.png | Archive support figure for MTUS country-level external-validity variation.P3-F10 | results/archive_support_figures/phase3_fig10_sample_size_vs_delta.png | Archive support figure for sample-size sensitivity in external-validity runs.P3-F11 | results/archive_support_figures/phase3_fig11_a1_country_ci_forest.png | Archive support figure for country-level uncertainty without sign reversal.P3-F12 | results/archive_support_figures/phase3_fig12_b1_distribution_boxplots.png | Archive support figure for grouped full-run distributions across countries.results/phase4plus_figures/phase4plus_fig1_delivery_status_matrix.pngORIGINAL_FIGURES/PHASE4PLUS_results_figures/phase4plus_fig1_delivery_status_matrix.pnglive summary jsonsresults/phase4plus_figures/phase4plus_fig2_cross_country_bridge.pngORIGINAL_FIGURES/PHASE4PLUS_results_figures/phase4plus_fig2_cross_country_bridge.pngresults/next_stage_cross_country_activity_master/summary.jsonresults/phase4plus_figures/phase4plus_fig3_mtus_wave1_country_delta.pngORIGINAL_FIGURES/PHASE4PLUS_results_figures/phase4plus_fig3_mtus_wave1_country_delta.pngresults/next_stage_cross_country_activity_master/summary.jsonresults/phase4plus_figures/phase4plus_fig4_groupb_axes.pngORIGINAL_FIGURES/PHASE4PLUS_results_figures/phase4plus_fig4_groupb_axes.pngresults/next_stage_groupb_master/summary.jsonresults/phase4plus_figures/phase4plus_fig5_extension_support.pngORIGINAL_FIGURES/PHASE4PLUS_results_figures/phase4plus_fig5_extension_support.pngresults/next_layer_extension_summary/summary.jsonresults/phase4plus_figures/phase4plus_fig6_evidence_gate.pngORIGINAL_FIGURES/PHASE4PLUS_results_figures/phase4plus_fig6_evidence_gate.pngresults/evidence_gate_decision/summary.jsonresults/anchor_figures/figure6_transition_analysis.pngORIGINAL_FIGURES/ANCHOR_results_figures/figure6_transition_analysis.pngresults/figures/figure6_transition_analysis.pngresults/anchor_figures/figure8_input_info_effect.pngORIGINAL_FIGURES/ANCHOR_results_figures/figure8_input_info_effect.pngresults/figures/figure8_input_info_effect.pngresults/anchor_figures/figure10_dimension_importance.pngORIGINAL_FIGURES/ANCHOR_results_figures/figure10_dimension_importance.pngresults/figures/figure10_dimension_importance.pngresults/project_figures/figure1_a_class_performance.pngORIGINAL_FIGURES/PROJECT_results_figures/figure1_a_class_performance.pngresults/figures/figure1_a_class_performance.pngresults/project_figures/figure2_b_class_heatmap.pngORIGINAL_FIGURES/PROJECT_results_figures/figure2_b_class_heatmap.pngresults/figures/figure2_b_class_heatmap.pngresults/project_figures/figure3_e2b_waterfall.pngORIGINAL_FIGURES/PROJECT_results_figures/figure3_e2b_waterfall.pngresults/figures/figure3_e2b_waterfall.pngresults/project_figures/figure4_e1_model_comparison.pngORIGINAL_FIGURES/PROJECT_results_figures/figure4_e1_model_comparison.pngresults/figures/figure4_e1_model_comparison.pngresults/project_figures/figure5_bootstrap_ci.pngORIGINAL_FIGURES/PROJECT_results_figures/figure5_bootstrap_ci.pngresults/figures/figure5_bootstrap_ci.pngresults/project_figures/figure7_imputation_robustness.pngORIGINAL_FIGURES/PROJECT_results_figures/figure7_imputation_robustness.pngresults/figures/figure7_imputation_robustness.pngresults/project_figures/figure9_unified_model_comparison.pngORIGINAL_FIGURES/PROJECT_results_figures/figure9_unified_model_comparison.pngresults/figures/figure9_unified_model_comparison.pngresults/project_figures/figure11_deep_transition.pngORIGINAL_FIGURES/PROJECT_results_figures/figure11_deep_transition.pngresults/figures/figure11_deep_transition.pngresults/project_figures/figure12_order_vs_bag.pngORIGINAL_FIGURES/PROJECT_results_figures/figure12_order_vs_bag.pngresults/figures/figure12_order_vs_bag.pngresults/project_figures/figure13_role_timelines.pngORIGINAL_FIGURES/PROJECT_results_figures/figure13_role_timelines.pngresults/figures/figure13_role_timelines.pngresults/project_figures/figure14_role_comparison.pngORIGINAL_FIGURES/PROJECT_results_figures/figure14_role_comparison.pngresults/figures/figure14_role_comparison.pngresults/project_figures/supplementary_fig1_deep_models_by_income.pngORIGINAL_FIGURES/PROJECT_results_figures/supplementary_fig1_deep_models_by_income.pngresults/figures/supplementary_fig1_deep_models_by_income.pngresults/project_figures/supplementary_fig2_multitask_comparison.pngORIGINAL_FIGURES/PROJECT_results_figures/supplementary_fig2_multitask_comparison.pngresults/figures/supplementary_fig2_multitask_comparison.pngresults/project_figures/supplementary_fig3_error_rate_by_group.pngORIGINAL_FIGURES/PROJECT_results_figures/supplementary_fig3_error_rate_by_group.pngresults/figures/supplementary_fig3_error_rate_by_group.pngresults/archive_support_figures/phase2_fig8_us_sample_size_delta_shift.pngORIGINAL_FIGURES/ARCHIVE_support_figures/phase2_fig8_us_sample_size_delta_shift.pngADVISOR_PACKAGE_PHASE1_2_3_20260403/ORIGINAL_FIGURES/PHASE2_results_phase2_figures/phase2_fig8_us_sample_size_delta_shift.pngresults/archive_support_figures/phase2_fig10_us_pooled_support_vs_delta.pngORIGINAL_FIGURES/ARCHIVE_support_figures/phase2_fig10_us_pooled_support_vs_delta.pngADVISOR_PACKAGE_PHASE1_2_3_20260403/ORIGINAL_FIGURES/PHASE2_results_phase2_figures/phase2_fig10_us_pooled_support_vs_delta.pngresults/archive_support_figures/phase3_fig1_a1_delta_by_country.pngORIGINAL_FIGURES/ARCHIVE_support_figures/phase3_fig1_a1_delta_by_country.pngADVISOR_PACKAGE_PHASE1_2_3_20260403/ORIGINAL_FIGURES/PHASE3_results_phase3_figures/phase3_fig1_a1_delta_by_country.pngresults/archive_support_figures/phase3_fig10_sample_size_vs_delta.pngORIGINAL_FIGURES/ARCHIVE_support_figures/phase3_fig10_sample_size_vs_delta.pngADVISOR_PACKAGE_PHASE1_2_3_20260403/ORIGINAL_FIGURES/PHASE3_results_phase3_figures/phase3_fig10_sample_size_vs_delta.pngresults/archive_support_figures/phase3_fig11_a1_country_ci_forest.pngORIGINAL_FIGURES/ARCHIVE_support_figures/phase3_fig11_a1_country_ci_forest.pngADVISOR_PACKAGE_PHASE1_2_3_20260403/ORIGINAL_FIGURES/PHASE3_results_phase3_figures/phase3_fig11_a1_country_ci_forest.pngresults/archive_support_figures/phase3_fig12_b1_distribution_boxplots.pngORIGINAL_FIGURES/ARCHIVE_support_figures/phase3_fig12_b1_distribution_boxplots.pngADVISOR_PACKAGE_PHASE1_2_3_20260403/ORIGINAL_FIGURES/PHASE3_results_phase3_figures/phase3_fig12_b1_distribution_boxplots.pngresults/phase4plus_figures/results/anchor_figures/results/project_figures/results/archive_support_figures/ORIGINAL_FIGURES/如果需要稳定引用,请优先使用 results/... 里的 canonical paths;ORIGINAL_FIGURES/... 更适合邮件、PPT 或单独抽图。