Paper LaTeX Review

Overview

Use this skill to turn a research paper PDF into a bilingual Chinese-English LaTeX review report. The report must prioritize the paper's own content, extract the truly important ideas, explain methods and representative comparison models, and assess reproducibility.

Source Policy

1. First analyze the PDF only.
2. Extract all claims that are available in the PDF: title, authors, venue if present, abstract, problem, contributions, method, equations, experiments, baselines, datasets, metrics, limitations, and reproducibility details.
3. Use web search only when the PDF lacks enough information for:

• representative model background or method details,
• current international status, SOTA, or mainstream research direction,
• definitions of important technical terms,
• code/data availability that is not stated in the PDF.

1. Clearly label every externally supplemented item as 外部补充 / External Supplement.
2. If a fact cannot be verified from the PDF or reliable external sources, write PDF 中未提供 / Not provided in the PDF or 未能确认 / Not verified instead of guessing.

Workflow

1. Read the PDF carefully and identify the paper type: theory, model architecture, dataset, benchmark, systems, empirical study, survey, or application paper.
2. Build a paper map:

• research question,
• motivation,
• task definition,
• proposed method,
• important equations,
• experiments and baselines,
• stated contributions,
• limitations and reproducibility evidence.

1. Decide which model comparisons are available from the PDF:

• baselines in experimental tables,
• models named in related work,
• ablations that represent prior variants,
• closest competing method discussed by the authors.

1. Explain representative models with method-level detail when possible:

• core idea,
• pipeline,
• objective or key equations,
• why it is used as a comparison,
• difference from the target paper,
• strengths and weaknesses.

1. Generate the LaTeX report using the structure below.
2. Preserve uncertainty and source boundaries. Do not make the report sound more certain than the evidence allows.

LaTeX Output Requirements

Return a single compilable .tex document unless the user asks for a full project. Use XeLaTeX-compatible Chinese support.

Use this default preamble:

latex
\\documentclass[11pt]{article}
\\usepackage[UTF8]{ctex}
\\usepackage{geometry}
\\usepackage{amsmath, amssymb, amsfonts}
\\usepackage{booktabs}
\\usepackage{enumitem}
\\usepackage{hyperref}
\\usepackage{xcolor}
\\geometry{margin=1in}
\\hypersetup{colorlinks=true, linkcolor=blue, urlcolor=blue, citecolor=blue}
\\title{论文精读报告 / Bilingual Paper Review}
\\author{Generated by ChatGPT}
\\date{\\today}

Use bilingual section titles in the format 中文 / English. Put Chinese first and English second.

Required Report Structure

Always include these sections unless the PDF makes a section impossible; if impossible, include the section and mark missing information.

latex
\\section{论文信息 / Paper Metadata}
\\section{核心重要总结 / Core Importance Summary}
\\section{摘要 / Summary}
\\section{问题背景与动机 / Problem Background and Motivation}
\\section{创新点 / Contributions and Innovations}
\\section{方法解释 / Method Explanation}
\\section{关键公式与符号 / Key Equations and Notation}
\\section{代表模型与对比模型来源 / Representative and Compared Models}
\\section{实验结果 / Experimental Findings}
\\section{重要名词解释 / Key Term Glossary}
\\section{国际现状分析 / International Research Status Analysis}
\\section{可复现性分析 / Reproducibility Analysis}
\\section{局限性与风险 / Limitations and Risks}
\\section{适合读者与使用场景 / Suitable Readers and Use Cases}
\\section{结论 / Conclusion}

Section Guidance

核心重要总结 / Core Importance Summary

Include:

• one-sentence Chinese summary,
• one-sentence English summary,
• 3分钟读懂 / Three-minute version,
• why the paper matters,
• who should read it.

摘要 / Summary

Do not merely translate the abstract. Summarize the real paper content:

• problem,
• method,
• main result,
• significance,
• caveats.

创新点 / Contributions and Innovations

For each innovation, include:

• Chinese explanation,
• English explanation,
• what prior limitation it addresses,
• evidence from the paper,
• whether the innovation is conceptual, architectural, algorithmic, empirical, dataset-related, or systems-related.

方法解释 / Method Explanation

Use two layers:

• 直觉解释 / Intuitive Explanation: explain as if to a technical reader new to the topic.
• 技术解释 / Technical Explanation: explain modules, pipeline, objective, training/inference, and implementation details.

When the paper has a model architecture, describe inputs, outputs, intermediate representations, training objective, and inference process.

关键公式与符号 / Key Equations and Notation

For each important equation:

• reproduce or paraphrase the equation in LaTeX,
• define every symbol,
• explain the equation's role,
• explain intuition,
• explain how it connects to the method.

If equations are embedded as images and cannot be reliably extracted, state that formulas need manual verification from the PDF image.

代表模型与对比模型来源 / Representative and Compared Models

Start with a source table:

latex
\\begin{tabular}{p{0.22\\linewidth}p{0.30\\linewidth}p{0.38\\linewidth}}
\\toprule
模型 / Model & 来源 / Source & 为什么重要 / Why It Matters \\\\
\\midrule
...
\\bottomrule
\\end{tabular}

Classify sources as:

• 实验表格 baseline / experimental baseline,
• Related Work,
• 作者讨论 / author discussion,
• 外部补充 / external supplement.

For each representative model, include:

latex
\\subsection{代表模型：MODEL / Representative Model: MODEL}
\\paragraph{核心思想 / Core Idea}
...
\\paragraph{方法流程 / Method Pipeline}
...
\\paragraph{关键公式或目标函数 / Key Equation or Objective}
...
\\paragraph{与本文方法的区别 / Difference from This Paper}
...
\\paragraph{优缺点 / Strengths and Weaknesses}
...

重要名词解释 / Key Term Glossary

Explain domain terms that are central to understanding the paper. For each term:

• Chinese name,
• English term,
• plain-language explanation,
• role in this paper,
• common confusion or related terms when useful.

国际现状分析 / International Research Status Analysis

First use the PDF's related work and citations. If insufficient, use web search and label additions as external supplements. Cover:

• mainstream research directions,
• how this paper fits into the field,
• whether the method is incremental, representative, or potentially paradigm-shifting,
• comparison with current international practice or SOTA when verifiable,
• likely future directions.

可复现性分析 / Reproducibility Analysis

Use this rubric:

latex
\\begin{tabular}{p{0.25\\linewidth}p{0.18\\linewidth}p{0.47\\linewidth}}
\\toprule
维度 / Dimension & 判断 / Assessment & 依据 / Evidence \\\\
\\midrule
代码 / Code & ... & ... \\\\
数据 / Data & ... & ... \\\\
超参数 / Hyperparameters & ... & ... \\\\
训练成本 / Training Cost & ... & ... \\\\
评测协议 / Evaluation Protocol & ... & ... \\\\
依赖与环境 / Dependencies & ... & ... \\\\
总体复现难度 / Overall Difficulty & ... & ... \\\\
\\bottomrule
\\end{tabular}

Use labels: 容易 / Easy, 中等 / Medium, 困难 / Hard, 无法判断 / Unknown.

Style Rules

• Be concise but information-dense.
• Prefer faithful technical explanation over promotional language.
• Use bilingual content throughout, not only translated headings.
• Keep equations in valid LaTeX syntax.
• Do not invent citation details.
• If using external information, include a short 信息来源说明 / Source Note subsection explaining what was PDF-based and what was externally supplemented.
• If the user asks for only a subset, still follow the same source policy and bilingual style.