by gulucaptain
Claude / Codex skill for drafting abstracts, introductions, methods and experiments in the style of CVPR/ICCV/NeurIPS best papers, distilled from 60 award-winning papers.
# Add to your Claude Code skills
git clone https://github.com/gulucaptain/Best-Paper-Writing-SkillGuides for using ai agents skills like Best-Paper-Writing-Skill.
See how Best-Paper-Writing-Skill compares with popular alternatives.
Best-Paper-Writing-Skill is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by gulucaptain. Claude / Codex skill for drafting abstracts, introductions, methods and experiments in the style of CVPR/ICCV/NeurIPS best papers, distilled from 60 award-winning papers. It has 53 GitHub stars.
Best-Paper-Writing-Skill's catalog security scan is still queued. You can run an instant dependency and prompt-injection check now with the "Scan for vulnerabilities" button above.
Clone the repository with "git clone https://github.com/gulucaptain/Best-Paper-Writing-Skill" and add it to your Claude Code skills directory (see the Installation section above).
Best-Paper-Writing-Skill is primarily written in TeX. It is open-source under gulucaptain on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh Best-Paper-Writing-Skill against similar tools.
No comments yet. Be the first to share your thoughts!
Available after the catalog security scan passes.
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM is an informational directory and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
The deep catalog scan for this skill is still queued. Run an instant dependency check now instead.
生成最佳论文风格的高质量论文底稿:摘要、引言、方法、实验与配图
🏆 写作规律提炼自 60 篇 CVPR / ICCV / NeurIPS 获奖论文,一个 Skill 同时支持 Claude 与 Codex
🌐 项目主页 · 中文 · English · 论文清单 · 更新日志
输入你的研究(填表或直接给摘要),输出一份接近获奖论文写法的稿子:
check_figures.py 检查图注长度、首句、引用、矢量图与占位图report.md:核心主张、主张-证据表、待补充清单,以及留给 Limitations 的要点[TODO]通用 AI 的稿子彼此相似:先铺背景,用形容词代替机制,并排罗列组件,句尾总带一个"……,making X possible"。获奖论文不是这样写的。
| 写法 | 通用 AI 草稿 | 获奖论文 |
|---|---|---|
| 句尾分词短语(", making …") | 约 32% 的句子 | 约 8% |
| 开头 | 泛泛的背景("has attracted increasing attention") | 具体的场景或前人工作 |
| 缺口 | 一句"现有方法有局限" | 什么条件下失败、为什么失败 |
| 组件 | A 做 …… B 做 …… C 做 ……(并排) | 第二个组件解决第一个留下的问题 |
| 局限 | 常写进摘要 | 留给 Limitations 一节 |
规则来自对语料的逐句标注与统计,每条规则都有上下限,而不是一味向"平均值"靠拢。
📝 填表 / 📄 贴摘要 → 🔍 补全缺项 → 🧭 判断类型 → ✍️ 参照范例起草 → ✅ 自检 → 📦 论文稿 + report.md
用浏览器打开 skills/bestpaper-writing/assets/intake_form.html(离线可用,自动保存草稿,可导出和再次导入 JSON)。必填只有 5 项,其余按需补充,补充得越详细,底稿质量越高:
| 档位 | 内容 |
|---|---|
| ⚙️ 基本设置 | 会议 / 年份 / 页数(下拉)、需要写哪些部分(勾选) |
| 🔴 必填(5) | 一句话说清这篇工作、任务、现有方法在什么情况下不行、方法组成、头条结果 |
| ⭐ 推荐(8) | 现有最强方法、失败原因、核心观察、最相近工作、具体画面、效率优势、反直觉之处、团队内部的叫法;缺了 skill 会先追问一次 |
| 🟢 补充(折叠) | 额外优势与应用、论文类型、矛盾类型、次要结果、开源内容、局限、其他材料、refs.bib |
| 🔵 方法与实验(折叠) | 一个直接粘贴材料的大框;需要时再逐项填写符号、组件、目标、数据集、指标、对比方法、结果表、消融 |
把导出的 JSON 发给 agent:
用 Best-Paper-Writing-Skill,根据这份信息表 JSON 写出论文的 Title、Abstract、Introduction、Methods 和 Experiments。
skill 不会盲目重写。它先把摘要拆回上面的字段,告诉你哪些已有、哪些模糊、哪些缺失,一次性追问缺项,再写出完整底稿,并给出摘要的修改对照表(原句 → 改后句 → 理由),你可以逐条接受或拒绝。你的数字和术语原样保留;摘要里没有的数字、基线和引用一律留作 [TODO]。
用 Best-Paper-Writing-Skill,根据这段摘要写出论文的 Title、Abstract、Introduction、Methods 和 Experiments。
也可以直接用文字描述工作。输出为 Markdown,或直接写进 LaTeX 模板(如 CVPR author kit)。
git clone https://github.com/gulucaptain/Best-Paper-Writing-Skill.git
cd Best-Paper-Writing-Skill
./install.sh claude # Claude Code → ~/.claude/skills/
./install.sh codex # Codex → ~/.agents/skills/
./install.sh all # 两个都装;加 --project 则装到当前项目目录
| 平台 | 方式 |
|---|---|
| 🟣 Claude Code | ./install.sh claude,或手动复制 skills/bestpaper-writing/ 到 ~/.claude/skills/ |
| 🟣 Claude.ai / Cowork | 把 bestpaper-writing 文件夹打成 zip,在 Skills 页面上传 |
| 🟢 Codex | ./install.sh codex,或手动复制到 ~/.agents/skills/;用 $bestpaper-writing 显式调用 |
两个平台使用同一种 SKILL.md 格式。装好后若看不到,重启 agent。辅助脚本需要 Python 3。
Best-Paper-Writing-Skill/
├── README.md / README_en.md
├── PAPER_LIST.md # 60 篇论文清单
├── data/ # papers.csv · figure_stats.json · figure1_types.csv
├── template/CVPR2027/ # 官方 CVPR 2027 LaTeX 模板
├── tools/ # 语料配图统计脚本
├── install.sh # 一键安装
└── skills/bestpaper-writing/
├── SKILL.md # 入口:工作流与硬性规则
├── references/ # 摘要、引言、方法、实验、配图、路由、风格规则、语料统计
│ └── exemplars/ # 段落与句式范例库
├── scripts/ # check_draft.py · check_figures.py · find_exemplars.py · intake_to_brief.py · style_metrics.py
├── assets/ # intake_form.html · figures/(TikZ 样式、teaser/总览/对比网格模板、matplotlib 样式)
├── evals/ # 留出测试集与盲评标准
└── agents/openai.yaml # Codex 界面配置
2021 至 2025 年 CVPR、ICCV、NeurIPS 的最佳论文、最佳学生论文与荣誉提名。
| 会议 | 篇数 |
|---|---|
| CVPR | 22 |
| ICCV | 12 |
| NeurIPS | 26 |
类型分布:方法改进 26 · 新能力 7 · 数据集 6 · 测量 5 · 机理分析 5 · 理论 4 · 规模规律 4 · 再检验 3。完整清单见 PAPER_LIST.md。
🔬 关于留出测试:风格规则由其中 55 篇拟合。另外 5 篇(均为 2024 年)从范例库中移除,只用来测试,避免规则在自己的测试集上调参。
references/exemplars/ 中的论文短句仅作教学评论引用并注明出处,版权归原作者,请勿复制进新论文。工作流与模板基于 Master-cai/Research-Paper-Writing-Skills(MIT),其写作知识来自彭思达的学习笔记。详见 NOTICE.md。
欢迎通过 Issue 反馈问题、补充语料中缺少的领域(NLP、机器人、系统等),或分享你用它写出的稿子。如果它对你有帮助,欢迎点一个 ⭐ Star。