菜单
Use when the research task is in math — optimisation, algorithms, solvers and formal or symbolic reasoning — at study design, analysis or writing. Draw the field's plain two-curve figure against the algorithm's own cost counter, before any richer diagnostic
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/tangxiangru/AutoR --skill math-canonical-curve-on-the-cost-counter命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Use when writing results and discussion, and when a task or a reviewer asks why an effect happens rather than whether it does. Covers the difference between reporting an effect and accounting for it, and what a mechanism claim needs behind it.
Use when the supplied data looks synthetic, degraded, incomplete or wrong, and whenever you are tempted to reframe the study around what you found about the inputs, the harness or the evaluation. Covers what to do with a real data problem without losing the study.
Use when planning figures, at study design and again before writing. Covers the figures a paper in this field is expected to contain, why an original figure does not substitute for a standard one, and how to decide what to draw first.
基于 SOC 职业分类
正在显示 SKILL.md
| name | math-canonical-curve-on-the-cost-counter |
| description | Use when the research task is in math — optimisation, algorithms, solvers and formal or symbolic reasoning — at study design, analysis or writing. Draw the field's plain two-curve figure against the algorithm's own cost counter, before any richer diagnostic |
In algorithms and optimisation the standard result artifact is one plain figure: the field's headline scalar on the ordinate against the algorithm's own implementation-independent cost counter on the abscissa, with every competing method on the same axes as an additional curve or a labelled horizontal reference line. Draw that figure first, alone, in its own panel. Folding it into an eleven-series multi-panel diagnostic destroys it -- a reader has to be able to see one curve sitting below another.
Pick the abscissa another group can reproduce on different hardware: iteration index, oracle or function evaluations, node expansions, sample count, instance size, number of agents or items. Wall-clock time is a legitimate second figure and never a substitute for the first; on a shared or loaded machine it is also the axis a referee discounts. For residuals, duality gaps and objective gaps use a logarithmic ordinate across the full decade span, and draw the target tolerance as an annotated horizontal line.
State the headline scalar in the paper's own unit on a three-point scale: prior method, your method, and the oracle, human or ideal reference where the field defines one. If the available instances are smaller than the benchmark the claim concerns, run the reduced set anyway, report the achieved-versus-target ratio explicitly, and support it with a scaling curve across instance size. An arm deferred in a limitations bullet is scored as an arm never run.
Math's headroom is entirely in the quantitative/figure criteria (71% of weight, agent mean 21/100), and the two lowest scores in the discipline (5 and 3) were both this exact failure: the canonical two-curve plot was produced but buried inside an eleven-method multi-panel composite (5/100 on a 0.30-weight item), and a progress curve was drawn against wall-clock instead of the algorithm's own iteration counter (3/100). It also converts the measured 'scoped out in Limitations' loss into a reduced-scale run with a stated achieved-versus-target ratio, which scored 25 rather than 0 the one time it was done.