
Paper Plot Skills
SkillPlotting skills for academic papers: bar, line, scatter, radar and other styles distilled from 9 real paper figures. Plot your own data in a chosen style as 300 dpi figures, or reproduce a matplotlib script from a paper screenshot. Needs local Python and matplotlib.
Paper Plot Skills
这里是我自己在用的,一套用于复现和生成学术论文图表的 AI Skills 工具箱。
我仔细选取了自己阅读过论文里绘图风格很有参考价值的各种图表,具有很强的参考意义,能极大程度减少绘图时候的重复工作。
这个仓库只是提供了一个起始点,从 9 张真实论文图表中提炼出系统化风格参数,支持按风格填数据和从图片复现两种使用方式。
这些skills会随着你的复现过程不断优化。
如果对你有用,麻烦点一个star🌟,鼓励我们不断优化这个仓库!
🎨 预置风格一览
| 风格 | 类型 | 来源论文 | 关键特征 |
|---|---|---|---|
[bar_paired_delta](#bar_paired_delta) | 柱状图 | MemEvolve | 配对柱 + 增益箭头,serif 字体 |
[bar_grouped_hatch](#bar_grouped_hatch) | 柱状图 | SPICE | 分组柱 + 斜线填充主方法,柱顶数值 |
[line_confidence_band](#line_confidence_band) | 折线图 | Self-Distillation | EMA 平滑 + 置信区间阴影,LaTeX 字体 |
[line_training_curve](#line_training_curve) | 折线图 | DAPO | 垂直断点线 + 水平参考线,sans-serif |
[line_loss_with_inset](#line_loss_with_inset) | 折线图 | SiameseNorm | L 形 spine + 轴端箭头 + 右侧 zoom inset |
[scatter_tsne_cluster](#scatter_tsne_cluster) | 散点图 | MemGen | t-SNE 聚类 + 圆角彩色注释框,点线网格 |
[scatter_broken_axis](#scatter_broken_axis) | 散点图 | Meta-Harness | 折断 X 轴双面板,多 marker 类型 |
[radar_dual_series](#radar_dual_series) | 雷达图 | DoRA | 正八边形虚线同心网格,双方法对比 |
Skills
| Skill | 说明 | 触发方式 |
|---|---|---|
| plot-from-data | 选择上方任意风格,填入你的数据,生成 dpi=300 论文图 | "用 bar_grouped_hatch 风格画我的数据" |
| plot-from-image | 上传论文截图,自动分析比例/字体/配色并复现为 matplotlib 脚本 | "帮我复现这张图" |
原图与复现总览
| 图名 | 原图 | 复现图 |
bar_memevolve | ![]() | ![]() |
bar_spice | ![]() | ![]() |
line_selfdistill_train | ![]() | ![]() |
line_selfdistill_scale | ![]() | ![]() |
line_aime | ![]() | ![]() |
line_loss_inset | ![]() | ![]() |
scatter_tsne | ![]() | ![]() |
scatter_break | ![]() | ![]() |
radar_dora | ![]() | ![]() |
classwise_iou | ![]() | ![]() |
其中 classwise_iou 是直接走 plot-from-image 流程新增复现的案例:
输入图片:
originals/classwise_iou.png
复现脚本:plot-from-image/scripts/classwise_iou_table.py
风格图库 · Gallery
柱状图 Bar Charts
bar_paired_delta — 配对增益柱
来源:MemEvolve: Meta-Evolution of Agent Memory Systems
serif 字体,配对柱(baseline vs method),箭头标注增益,Y 轴各子图独立
参数文档:plot-from-data/references/bar_paired_delta.md· 脚本:plot-from-data/scripts/bar_memevolve.py
| 原图 | 复现 |
![]() | ![]() |
bar_grouped_hatch — 分组斜线填充柱
来源:SPICE: Self-Play In Corpus Environments
LaTeX serif,分组柱 + 主方法白色斜线填充,柱顶数值(最优加粗),开口 L 形 spine
参数文档:plot-from-data/references/bar_grouped_hatch.md· 脚本:plot-from-data/scripts/bar_spice.py
| 原图 | 复现 |
![]() | ![]() |
折线图 Line Charts
line_confidence_band — 置信区间训练曲线
来源:Reinforcement Learning via Self-Distillation
LaTeX Computer Modern serif,EMA 平滑主线,浅色置信区间fill_between,SDPO 加粗图例
参数文档:plot-from-data/references/line_confidence_band.md· 脚本:plot-from-data/scripts/line_selfdistill.py
| 原图 | 复现 |
![]() | ![]() |
line_training_curve — 垂直断点 + 水平参考线
来源:DAPO: An Open-Source LLM RL System at Scale
sans-serif,四边框,朝外刻度,水平参考线(独立蓝色),两条垂直断点虚线(与曲线同色)
参数文档:plot-from-data/references/line_training_curve.md· 脚本:plot-from-data/scripts/line_aime.py
| 原图 | 复现 |
![]() | ![]() |
line_loss_with_inset — L 形 spine + 局部放大 inset
来源:SiameseNorm: Breaking the Barrier to Reconciling Pre/Post-Norm
LaTeX serif,L 形 spine + 轴端箭头,虚线放大框,黑色虚线连接右侧独立 inset 子图
参数文档:plot-from-data/references/line_loss_with_inset.md· 脚本:plot-from-data/scripts/line_loss_inset.py
| 原图 | 复现 |
![]() | ![]() |
散点图 Scatter Plots
scatter_tsne_cluster — t-SNE 聚类分布
来源:MemGen: Weaving Generative Latent Memory for Self-Evolving Agents
LaTeX serif,7 类聚类,圆角注释框(统一深灰边 + 聚类色底),浅灰点线网格,四边框
参数文档:plot-from-data/references/scatter_tsne_cluster.md· 脚本:plot-from-data/scripts/scatter_tsne.py
| 原图 | 复现 |
![]() | ![]() |
scatter_broken_axis — 折断 X 轴散点图
来源:Meta-Harness: End-to-End Optimization of Model Harnesses
sans-serif 粗体标签,双面板折断 X 轴(0-50k | 115k/200k),多 marker(★ ○ △ ◆ × ○),折断符仅底边
参数文档:plot-from-data/references/scatter_broken_axis.md· 脚本:plot-from-data/scripts/scatter_break.py
| 原图 | 复现 |
![]() | ![]() |
雷达图 Radar Chart
radar_dual_series — 双方法多维对比
来源:DoRA: Weight-Decomposed Low-Rank Adaptation
sans-serif,正八边形虚线同心网格,DoRA 深绿粗线 vs LoRA 蓝色细线,数值标注白底,图例左上
参数文档:plot-from-data/references/radar_dual_series.md· 脚本:plot-from-data/scripts/radar_dora.py
| 原图 | 复现 |
![]() | ![]() |
plot-from-image 示例
classwise_iou — 类别级结果表
来源:用户上传论文截图 https://github.com/Trae1ounG/paper-plot-skills/issues/1
纯表格式布局,双行结果 + 强弱高亮底色,按图片比例与文字排布直接复现
输入图片:originals/classwise_iou.png· 脚本:plot-from-image/scripts/classwise_iou_table.py
| 原图 | 复现 |
![]() | ![]() |
Star History
관련 리소스

Scientific Figure Making
SkillPublication-quality matplotlib plotting skill from a Yale PhD student: bar charts, trends, heatmaps and multi-panel figures with consistent fonts, colors and export settings, plus real plotting scripts from papers published in venues like Nature Machine Intelligence. Needs Python.
ChenLiu-1996/figures4papers8.6k
Math Modeling Skill
SkillA three-stage workflow for math modeling contests like MCM/ICM and CUMCM: analysis, solving and plotting in Python or MATLAB, and a Word (optionally LaTeX) paper, with stage-check subagents and OpenAlex plus AnySearch literature search. Generated papers are for reference only.
XiaoMaColtAI/math-modeling-skill1.9k
Scientific Agent Skills
Skill166 research skills maintained by K-Dense (formerly Claude Scientific Skills), spanning bioinformatics, cheminformatics, clinical research, materials and physics, including unified access to 78 public databases such as PubChem and UniProt. Works with Claude Code, Codex and Cursor.
K-Dense-AI/scientific-agent-skills51.2k
Statistical Analysis
SkillGuides agents to the right statistical method for research data: t-tests, ANOVA, chi-square, regression and Bayesian analysis, checking assumptions, computing effect sizes and power, and reporting results in APA format. Python-based (pingouin, statsmodels); by K-Dense.
K-Dense-AI/scientific-agent-skills51.2k
Datasette
CLIPublishes SQLite databases as browsable, queryable web pages and JSON APIs, with SQL queries, metadata and plugins, and deploys to Cloud Run and other platforms with datasette publish. Great for data exploration and data journalism. Needs Python 3.10+.
simonw/datasette11.5k
PaperOrchestra
SkillThe five-agent PaperOrchestra pipeline (outline, figures, literature review, section writing, polishing) as skills: turns research ideas, experiment logs and a conference template into a LaTeX paper and checks citations via Semantic Scholar. Needs Python; verify the draft yourself.
Ar9av/PaperOrchestra682



















