cumcm-b-problem

Follow the standard structure: one-page abstract with keywords, problem restatement, problem analysis, assumptions, notation, per-question model building and solving, verification, and evaluation. Each question should present a unified model equation block, results with interpretation, and validation.

Match the method to the problem structure: exact tests and regression for statistics, analytic geometry for positioning, enumeration or integer programming for small discrete decisions, and Monte Carlo for verification. Reserve GA, PSO, or machine learning for genuinely non-convex or large-scale bottlenecks.

Use at least two of three verification types: theoretical checks (limits, units, degenerate cases), independent numerical cross-checks (Monte Carlo or a second algorithm), and data validation (residuals, confidence intervals). Sensitivity analysis should target real strategy-switching thresholds, not mechanical ±5% perturbations.

No, the guidance is calibrated specifically to Problem B patterns from recent years. The skill explicitly warns against mechanically applying its B-problem experience to A or C problems, which have different structures and judging expectations.

Run XeLaTeX at least twice on example.tex, then inspect the log for Overfull boxes, undefined references, and citation warnings. The validate_project.py script automates structure checks, page-count verification against the 25-29 page target, and first-page abstract detection.