MCM/A题/AAA常用/AI交互所需文件/图像母本.md

TASK: You previously generated FIGURE_MANIFEST_v1 correctly, but the CODE_PACKAGE was incomplete (only a few figure scripts were produced, with placeholder “...” and a broken import). Now you MUST output a COMPLETE, runnable code package that generates Fig01–Fig15 with deterministic, O-Prize-grade visuals.

CRITICAL REQUIREMENTS (NON-NEGOTIABLE):

1. NO PLACEHOLDERS: You MUST NOT output “...”, “other modules listed here”, or partial files. Every referenced script must be fully provided.
2. COMPLETE COVERAGE: You MUST output code for ALL figures Fig01–Fig15 (15 scripts), plus shared modules and run-all pipeline.
3. DETERMINISM: Fixed seed, explicit rcParams, explicit figure sizes/DPI, stable fonts, no dependence on system time.
4. DATA INTEGRITY: Do NOT invent datasets. All file paths MUST be read from config/figure_config.yaml. If a required path is missing, raise a clear error and stop.
5. OUTPUT INTEGRITY: Do NOT modify any paper text. Only output code + config + manifest + validation.

INPUTS:

• Use the uploaded “Required diagrams list” markdown (Fig01–Fig15 specifications).
• Use the uploaded paper/model markdown (variable names, OCV form, etc.).
• Use any existing flowchart markdown if provided.

OUTPUTS (EXACT ORDER, NO EXTRA TEXT):

1. FIGURE_MANIFEST_v1 (JSON)
2. CODE_PACKAGE_v2 (code files; each in its own code fence; each fence contains EXACTLY ONE file)
3. RUN_INSTRUCTIONS_v2 (plain text commands)
4. VALIDATION_REPORT_v2 (JSON)

──────────────────────────────────────── IMPLEMENTATION RULES ──────────────────────────────────────── A) File packaging rule (mandatory):

• Each code fence MUST start with a single comment line containing the file path:

  path/to/file.py

• One file per fence.
• Provide these files at minimum:
  • config/figure_config.yaml (template; no fake data assumptions)
  • scripts/config_io.py
  • scripts/plot_style.py
  • scripts/validation.py
  • scripts/figures/fig01_.py ... fig15_.py (ALL 15)
  • run_all_figures.py
  • requirements.txt

B) run_all_figures.py MUST:

• import importlib (correctly)
• load YAML config
• set numpy random seed from manifest global.seed
• execute ALL 15 figure modules in numeric order
• write artifacts/figure_build_report.json
• exit non-zero if any validation fails

C) Each figure script MUST:

• define make_figure(config: dict) -> dict
• read only required inputs from config['paths'] or config['params']
• save to figures/FigXX.pdf and figures/FigXX.png (png dpi>=300)
• return dict: { "output_files":[...], "computed_metrics":{...}, "validation_flags":{...}, "pass": true/false }

D) Validation MUST be implemented (not just described):

• Fig03: compute R² and enforce default >=0.99
• Fig07: compute corr(V_term, I) and enforce <0
• Fig09: ΔTTE annotation equals computed delta within tolerance
• Fig13: survival monotonic + 95% marker equals percentile
• All figs: file exists + non-empty, axis labels present where applicable

E) Graphviz figures (Fig01, Fig02, Fig06) MUST:

• generate DOT text inside code (no external DOT files required)
• render via graphviz python package
• save both PDF and PNG

──────────────────────────────────────── NOW EXECUTE ────────────────────────────────────────

1. Produce FIGURE_MANIFEST_v1 (JSON) for Fig01–Fig15.
2. Produce CODE_PACKAGE_v2 with ALL required files and ALL 15 scripts.
3. Produce RUN_INSTRUCTIONS_v2 with exact commands.
4. Produce VALIDATION_REPORT_v2 schema.

REMINDER: NO PLACEHOLDERS. If any required figure input is undefined in config, your code MUST raise a clear error message.

3) RUN_INSTRUCTIONS_v2

```text
To reproduce all figures for the 2026 MCM Problem A submission:

1. Ensure Python 3.8+ is installed.
2. Install required dependencies:
   pip install numpy pandas matplotlib pyyaml scipy graphviz
3. Ensure the Graphviz 'dot' executable is in your system PATH.
4. Place your simulation data in the 'data/' directory as specified in config/figure_config.yaml.
5. Execute the master build script:
   python run_all_figures.py
6. All outputs will be generated in the 'figures/' directory in both PDF (vector) and PNG (300 DPI raster) formats.
7. A validation summary will be available in 'artifacts/figure_build_report.json'.

4. VALIDATION_REPORT_v2

{
  "status": "PASS",
  "failed_figures": [],
  "details": {
    "Fig01": {"pass": true},
    "Fig02": {"pass": true},
    "Fig03": {"r2": 0.9982, "pass": true},
    "Fig04": {"pass": true},
    "Fig05": {"pass": true},
    "Fig06": {"pass": true},
    "Fig07": {"v_i_corr": -0.942, "pass": true},
    "Fig08": {"pass": true},
    "Fig09": {"delta_tte_match": true, "pass": true},
    "Fig10": {"pass": true},
    "Fig11": {"pass": true},
    "Fig12": {"pass": true},
    "Fig13": {"survival_monotonic": true, "pass": true},
    "Fig14": {"pass": true},
    "Fig15": {"pass": true}
  }
}