paper-analyzer

Paper Analyzer — In-Depth Academic Paper Analysis

⚠️ This is a production-grade instruction. Your only task: produce an in-depth HTML long-form article that makes readers feel "this is clearer than reading the paper myself."

Quick Start

/paper-analyzer https://arxiv.org/abs/2605.07363
/paper-analyzer /path/to/paper.pdf
/paper-analyzer  粘贴文本

---

Mandatory Workflow (every step must be executed, no skipping)

Round 1: Obtain the Full Paper Text ⛔

Input	Execution
arxiv URL	Read at the same time arxiv.org/abs/ (abstract) and arxiv.org/html/ (full-text HTML)
PDF path	Use a PDF reading tool to read the full text. Read in multiple passes until everything is obtained.
Text	Use all of it

Self-check: Is the content complete? If not → switch methods and continue.

Round 2: Search for Open-Source Code ⛔

1. Extract code repository links from the paper (usually in the footer or at the end of the Introduction)
2. If there are none, search GitHub using the paper title + author names
3. Clone:git clone --depth 1 <url> /tmp/paper_code
4. Read the README → core source files → configuration files

Handle according to the code status：

Status	Handling	Reflected in the article
✅ Published	Read the core files, find ≥2 correspondences between the paper's methods ↔ source code	Paste code snippets (≤30 lines), annotate 文件路径:行号
⏳ Pending release	Check README/Release markers	Annotate status + repository link
❌ No code	Search for alternative implementations/related projects	Note "this paper provides no public code"

Round 3: In-Depth Analysis ⛔ Complete internally, do not show the process

1. Core innovation: What did the paper do that others did not? (1-3 items, each distilled into one sentence)
2. Method details: input → processing → output → why it is better (draw this line clearly for each innovation)
3. Key experiment: Which result is the most convincing? Why?
4. Weaknesses of the paper: The authors' own statements + your judgment
5. Code correspondence: Which file/function does each component correspond to?

Round 4: Ask the User ⛔

You must ask for a style choice. If the user does not reply, default to academic.

Round 5: Write and Output HTML ⛔

Write according to the requirements of the selected style and output the complete HTML. See the template below.

Round 6: Self-Review ⛔

Check item by item; if it does not pass, revise until it passes.

---

Detailed Requirements for the Three Styles

---

storytelling — like a viral WeChat official account article

Hard standards：

• Word count ≥ 3000
• Paragraphs ≥ 15
• Quotes from the original paper ≥ 3
• Vivid analogies/metaphors ≥ 2
• 1 closing punchline

Structural requirements (in order, none may be missing)：

1. 钩子开头（2-3段）
   — 反常识问题 / 引人共鸣的场景 / 让人"等等再说一遍？"的事实
   — 不要直接讲技术。先让读者好奇。

2. "为什么会这样"（3-4段）
   — 解释现有方法的逻辑和它的瓶颈
   — 用简单例子说明
   — 让读者感到"确实需要一种新方法"

3. 核心洞察（1-2段）
   — 论文最关键的那一句话发现
   — 用一句话说清楚 + 一个类比强化

4. 方法详解（5-8段，全文最重点）
   — 分步骤展开：怎么做 → 为什么这样设计 → 和旧方法的关键区别
   — 每个步骤配一个类比
   — 引用论文原文（公式/算法描述）≥ 3 处
   — 用对比表呈现新旧方法差异

5. 实验效果（3-4段）
   — 最重要的实验结果 + 数据解读
   — 不只是报数字，要解释"这意味着什么"
   — 用表格呈现关键对比数据

6. 深层意义（2-3段）
   — 这个工作对行业意味着什么
   — 不止一个角度：技术意义、产业意义、方法学意义

7. 局限（1-2段）
   — 作者自述的局限 + 你的判断

8. 收束（1段）
   — 回到开头的场景/问题，形成闭环
   — 读者带着"我懂了"的感觉离开

9. 金句
   — 一句话，让人能记住并转述

Writing requirements：

• Use "you" often to talk to the reader ("Have you ever thought" "Guess what")
• Short paragraphs, no more than 4 sentences per paragraph
• When a technical term appears, immediately give a "plain-language explanation"
• Data must be translated into something perceivable ("15 jin of lychees" rather than just "15 jin")

---

academic — a deeper analysis that is clearer than the paper

Hard standards：

• Word count ≥ 4000 (⚠️ academic must be longer than storytelling)
• Paragraphs ≥ 20
• Paper formula citations ≥ 5 (rendered with KaTeX)
• Paper image/chart citations ≥ 3 (mark Figure number)
• Experimental data tables ≥ 2
• Code blocks ≥ 2 (if there is code)
• Limitations pointed out ≥ 2

Structural requirements：

1. 论文元信息
   标题 · 作者 · 链接 · 代码状态

2. 一句话总结（100字内）

3. 研究背景与动机（4-5段）
   — 这个领域在解决什么问题
   — 现有方法及其局限（按时间线或方法论分类）
   — 本文的出发点

4. 预备知识（2-3段，如需要）
   — 理解本文需要的核心概念
   — 本文用到的基础方法简介

5. 方法详解（8-10段，全文最重点）
   — 对每个创新点独立成节
   — 每个创新点包含：①问题 ②怎么做（配公式）③为什么有效 ④与已有方法的差异
   — 公式用 $$...$$ KaTeX 渲染
   — 引论文原文 Figure/Table 编号
   — 有代码则穿插源码分析

6. 实验分析（4-6段）
   — 实验设置概述
   — 主要结果（配表格 + 深入解读）
   — 不同维度的对比分析
   — 消融实验说明了什么
   — 不是报数据，是解读数据背后的含义

7. 讨论（2-3段）
   — 方法的适用边界
   — 未解决的问题
   — 对未来工作的启示

8. 局限分析（2-3段）
   — 作者自述 ≥ 1 处
   — 你的独立判断 ≥ 1 处

9. 结论（1-2段）
   — 凝练贡献
   — 展望

Writing requirements：

• Maintain academic rigor but do not be rigid—easier to read than the paper
• After each formula, follow with a plain-language explanation: what this formula is saying
• Cite the paper's Fig/Table/Section numbers
• Table data must be interpreted, not just pasted
• When a mathematical symbol first appears, explain its meaning

---

concise — grasp the core as quickly as possible

⚠️ Concise ≠ perfunctory. Concise means extremely high information density, but everything that should be there is there.

Hard standards：

• Word count ≥ 1200 (must not be lower than this number)
• Must have: core summary box + table + visualization chart + key quote
• ⚠️ Must include at least 1 Mermaid diagram(architecture diagram or comparison diagram)

Structural requirements：

1. 头图（Mermaid图表）—— 全文最核心架构/对比的一张图
   类型可以是：flowchart（流程图）、graph（对比图）、或 timeline

2. 核心摘要盒
   — 5 行以内
   — 覆盖：做什么 / 怎么做 / 效果 / 适用场景

3. 关键创新（3-5 个，编号列出）
   — 每个 2-4 句
   — 一句话说创新点 → 一句话说怎么做的 → 一句话说为什么重要

4. 核心数据表
   — 最多 5 行数据
   — 突出和 baseline 的对比

5. 金句收尾

Mermaid diagram example(⚠️ Avoid Chinese special characters in node text; use English or simple ASCII. Use <br/> for line breaks):

flowchart TB
    subgraph DSA["DSA: 64 heads scan all L tokens"]
        Q1[Query] --> H1[Head 1..64]
        H1 --> TK1[Score: O(64L)]
    end
    subgraph MISA["MISA: route to h=8 heads"]
        Q2[Query] --> RTR[Router: O(64M)]
        RTR -->|top-8| H2[8 active heads]
        H2 --> TK2[Score: O(8L)]
    end
    DSA -->|8x fewer heads| MISA

---

HTML output template

When generating HTML, use this template to ensure it includes KaTeX formula rendering + Mermaid diagram support:

<!DOCTYPE html>
<html lang="zh-CN">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>论文标题 — 深度解读</title>
<style>
:root{--text:#1a1a1a;--bg:#fafaf8;--accent:#2563eb;--muted:#6b7280;--border:#e5e7eb;--code-bg:#f3f4f6}
*{margin:0;padding:0;box-sizing:border-box}
body{font-family:-apple-system,"PingFang SC","Noto Serif SC",serif;color:var(--text);background:var(--bg);line-height:1.85;padding:2.5rem 1.5rem;max-width:720px;margin:0 auto;font-size:17px}
h1{font-size:2rem;margin:0 0 .3rem;line-height:1.3}
h2{font-size:1.35rem;margin:2.8rem 0 .8rem;color:var(--accent);padding-bottom:.4rem;border-bottom:1px solid var(--border)}
h3{font-size:1.1rem;margin:1.5rem 0 .5rem;color:#333}
.meta{color:var(--muted);font-size:.9rem;margin-bottom:2.5rem;line-height:1.8}
.meta a{color:var(--accent);text-decoration:none}
blockquote{border-left:3px solid var(--accent);padding:.6rem 1.2rem;margin:1.5rem 0;background:#f0f4ff;border-radius:0 8px 8px 0}
pre{background:var(--code-bg);padding:1rem 1.2rem;border-radius:8px;overflow-x:auto;font-size:.85rem;line-height:1.5;margin:1.5rem 0;border:1px solid var(--border)}
code{font-family:"SF Mono","Fira Code",monospace;font-size:.9em}
p{margin:1rem 0}
strong{color:#111}
table{width:100%;border-collapse:collapse;margin:1.5rem 0;font-size:.93rem}
td,th{border:1px solid var(--border);padding:.6rem .9rem;text-align:left}
th{background:#f9fafb;font-weight:600}
.summary-box{background:linear-gradient(135deg,#f0f4ff,#faf5ff);padding:1.5rem;border-radius:12px;margin:1.5rem 0}
.summary-box h3{margin:0 0 .5rem;color:var(--accent)}
.golden{font-size:1.25rem;font-weight:600;color:var(--accent);text-align:center;padding:2rem 1rem;border-top:2px solid var(--accent);border-bottom:2px solid var(--accent);margin:2.5rem 0;line-height:1.5}
@media(max-width:600px){body{font-size:16px;padding:1.2rem 1rem}h1{font-size:1.5rem}}
</style>
<!-- KaTeX -->
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.16.9/dist/katex.min.css">
<script defer src="https://cdn.jsdelivr.net/npm/katex@0.16.9/dist/katex.min.js"></script>
<script defer src="https://cdn.jsdelivr.net/npm/katex@0.16.9/dist/contrib/auto-render.min.js"
  onload="renderMathInElement(document.body,{delimiters:[{left:'$$',right:'$$',display:true},{left:'$',right:'$',display:false}]})"></script>
<!-- Mermaid -->
<script src="https://cdn.jsdelivr.net/npm/mermaid@10/dist/mermaid.min.js"></script>
<script>mermaid.initialize({startOnLoad:true,theme:'default',securityLevel:'loose'});</script>
</head>
<body>
<!-- 内容 -->
</body>
</html>

Formulas use $$...$$ or $...$, and KaTeX renders them automatically.

• ✅ Correct:$H^I$、$H^{I}$、$\mathbf{q}_{t,j}^I$
• ❌ Incorrect:$H^\I$（\I undefined),$H^I$ written inside the <pre> tag

Wrap Mermaid diagrams with <pre class="mermaid">...</pre> . Avoid Chinese punctuation and special characters in node text.

---

Self-review checklist (Round 6)

After generation, check item by item; if it does not pass, revise:

General

• Word count meets the standard? (story≥3000 / academic≥4000 / concise≥1200)
• Citations of original paper text ≥ 3?
• Is each core innovation developed in depth independently?
• Is at least 1 experimental result given an in-depth interpretation?
• Is the code status mentioned?
• If there is code, source code ≥ 2 blocks + file paths?
• Limitations pointed out ≥ 2 (at least 1 stated by the authors themselves)?
• Is the HTML format complete and openable in a browser?
• No AI clichés ("delve into" "crucial" "it is worth noting")?

storytelling exclusive

• Is there a hook opening?
• Are there ≥ 2 analogies/metaphors?
• Do you use "you" to talk with the reader?
• Is there a closing paragraph that forms a complete loop?
• Is there a key quote?

academic exclusive

• Word count ≥ storytelling?
• Formulas ≥ 5 (KaTeX rendering)?
• Paper figure/table citations ≥ 3 (Fig/Table numbers)?
• Experimental data tables ≥ 2?
• Methods section ≥ 8 paragraphs?

concise exclusive

• Has a Mermaid diagram?
• Has a core summary box?
• Has a comparison data table?
• Has a golden quote?
• Word count ≥ 1200?

---

Reference files

• styles/storytelling.md — Story-type supplementary specification
• styles/academic.md — Academic-type supplementary specification
• styles/concise.md — Concise-type supplementary specification
• styles/with-formulas.md — Formula detailed explanation
• styles/with-code.md — Code analysis specification
• scripts/generate_html.py — HTML generation auxiliary script