Comprehensive guide for academic paper writing and formatting in English. Covers language issues, style guidelines, organization, and LaTeX/Word tools. Use when user asks about "academic writing", "paper format", "LaTeX", "research paper", "thesis writing", "manuscript preparation", or requests help with "abstract", "introduction", "related work", "conclusion" sections.
This skill provides comprehensive guidance for writing and formatting academic papers in English, based on common issues identified across 100+ papers. It covers language level problems, style issues, organization, and tools.
Language Level Issues
Voice and Pronoun Usage
CRITICAL: Use Active Voice in Abstract and Introduction
In abstract and contribution sections, prefer active voice over passive
Use "we" boldly - it represents the authors who proposed the model/algorithm
Do not use "this paper" repeatedly or passive voice unnecessarily
When to use "we":
In abstract and contribution summarization: "We propose...", "We design..."
When describing author actions and contributions
When NOT to use "we":
When describing the model itself (it's not about the authors)
When explaining model characteristics
Verwandte Skills
Example of WRONG usage: "The proposed framework uses intersections, so we can avoid noise influence." → The framework avoids noise, not "we"
Tense Rules
Use simple present and present perfect throughout (except Related Work)
Related Work: use past tense when describing previous work
Only experimental section may occasionally use past tense
Common Language Mistakes
DO NOT USE:
"can be", "may be" in abstract/contribution/conclusion when describing what you did
Use "is", "are" instead - you've done it, so state it directly
"We believe", "We think" - data/theory/experiments matter, not beliefs
"And" at the beginning of sentences (colloquial)
"very important" - use "vital", "fundamental", "paramount", "core issue"
USE CORRECTLY:
"cannot" (one word, NOT "can not")
"such as" (NOT "like" for examples)
"different from" (NOT "unlike")
"i.e." (NOT "namely", "aka")
"find" (NOT "find out")
"attempt to" (variation with "try to")
"that" in object clauses (don't omit it)
"as follows" (NOT "as following" or "as followed")
Punctuation and Formatting
NO Chinese punctuation: No book title marks (《》), no long dashes (——), no enumeration commas (、)
In LaTeX: quotes are `` '' (two backticks, two apostrophes)
Space after periods and commas: . Word NOT .Word
Space around parentheses: (word) NOT (word ) or ( word)
Sentences should be 20-30 words maximum for beginners
Split 50-word sentences into two 25-word sentences
Grammar Issues
Subject-verb agreement: check singular/plural
"Let something do" (NOT "does")
"training set", "test set" (NOT "testing set")
"prove" → use "verify" or "testify" instead (few things can be truly proven)
Avoid "mainly include", "mainly contain" - just list what's included
"if XXX, then XXX" - the "then" is unnecessary
Model Description Guidelines
Avoid subjective language in model descriptions
Do NOT use phrases like "we are interested in", "we focus on"
These add unnecessary subjectivity to technical descriptions
Example:
WRONG: "The distribution we are interested in is a conditional language model based on the input statement x."
RIGHT: "The distribution is a conditional language model based on the input statement x."
Subject Matching Errors
CRITICAL: Watch for dangling participles
WRONG: "When using XXXX to deal with the problem, the following steps are required."
"using" is done by humans, not by "the following steps"
RIGHT: "When XXXX is used to deal with the problem, the following steps are required."
Pronoun and Article Usage
In Related Work, referring to authors: use "authors" (plural), NOT "author" (singular)
Computer science papers typically have multiple authors
Generic references: use "users" or "a user" (NOT "the user" for generic reference)
Avoid redundant: "the two both datasets" → either "the two datasets" OR "both datasets"
Same for "the total number of the whole bikes" → "the total number of bikes"
Conference papers: "In Proceedings of XXX" (proceedings contain the papers)
References should include all authors (no "et al." in reference list)
Reference format must be consistent throughout
Authoritative Citation Sources:
DO NOT cite: Wikipedia, Baidu Baike, or other encyclopedias
DO cite: peer-reviewed conference papers, journals, books, monographs
Avoid arXiv papers IF published version exists (cite the published version)
Acceptable: websites for dataset sources or software repositories
Reference Formatting:
Journal papers: include volume, number, pages, year
Conference papers: generally no volume/number (unless published by journal)
Format: "In Proceedings of XXX" (papers are contained in proceedings)
Do NOT use [C][J][D] markers (these are Chinese format, NOT English)
Organization and Structure
Section Guidelines
Introduction:
Use "First", "Second", "Third" for enumeration
Clearly distinguish your method from baselines
Emphasize "the proposed method", "our proposed model"
Use correct verbs: "propose", "design", "develop", "study", "explore", "investigate"
Related Work:
Should have at least 20 references for full papers
Use past tense when describing previous work
Once you use past tense for a work, continue with past tense
Write everything in your own words - NO copying from cited papers
2/3 consecutive words repeated = plagiarism
1/2 words (not necessarily consecutive) = flagged as possible duplication
The Proposed Method:
Explain baselines appropriately, NOT in great detail
Don't derive standard formulas for baselines
Focus on YOUR contribution
Experiments:
Avoid "we can see XXX" for non-obvious phenomena
Use objective: "Figure 5 shows..."
Don't just describe trends - explain WHY trends occur
WRONG: "The accuracy first increases and then decreases."
RIGHT: "The accuracy increases from 70% to 85% as training epochs grow from 1 to 50, likely because the model learns more representative features. However, after 50 epochs, accuracy declines to 82%, suggesting overfitting on the training data."
For evaluation metrics: give full form if abbreviation used
Conclusion:
NOT a copy of Abstract
Write what you proposed, results obtained, what this demonstrates
Shows your hypothesis/design approach is correct
Visual Elements
Figures:
Must be clear, properly sized, correctly positioned
Labels should be readable
Use Python/Matlab/Mathematica (NOT Excel when possible)
Figure captions go BELOW the figure
Use vector formats (.pdf, .eps) for LaTeX (NOT .jpg, .png)
Tables:
Table captions go ABOVE the table
Font size should be one size smaller than body text
Formulas:
MUST have preceding explanation before presenting
MUST have following explanation after presenting
"where" is lowercase (NOT "Where" - it's part of the same paragraph)
Use standard mathematical notation
Use × or · for multiplication (NOT asterisk/snowflake)
Equation numbering: either all numbered or none numbered (be consistent)
Mathematical symbols (m, n, d, etc.) should be italicized in LaTeX
Titles and Headings
Use noun phrases or gerund phrases
Can include modifying adjectives/verbs
DO NOT start with verbs or adjectives
Level 1 and level 2 headings: format consistently
Either "Introduction" or "INTRODUCTION" throughout (not mixed)
Same for "The Proposed Method", "Experiments", etc.
Do NOT mix: "Introduction" then "RELATED WORK" then "Conclusion"
Document Layout
NO large whitespace gaps (5+ consecutive blank lines)
Paper folder naming: use meaningful names
BAD: "1", "paper", "doc"
GOOD: "The manuscript", "paper for AAAI", "ICML submission"
Mathematical Notation
Matrices: UPPERCASE (A, B, X)
Scalars: regular font
Vectors: lowercase (traditional ML: bold or subscript; neural networks: not strictly enforced)
Sets: UPPERCASE
Mathematical symbols must be italicized in LaTeX:
Variables like $m, n, d, i, j$
Example: "for $m$ users and $n$ items" (not "for m users")
Note: brackets, numbers, subscripts that are NOT variables should NOT be italic
Common Technical Mistakes
Common Confusions
"denote" vs "donate" (completely different meanings)
"the same as" NOT "the same with"
"compared to" NOT "compared with"
"because" or "as" NOT "since" (since means "now that" in modern English)
"XXX-based YYY" with hyphen (NOT "XXX based YYY")
"identical to" is a good alternative to "the same as"
Common Expression Errors
WRONG expressions to avoid:
"As can be seen from the results" → Use "As it can be seen from the results" OR "From the results, we can observe that"
"find out" → Use "find" alone (no need for "out")
"the two both datasets" → Use "the two datasets" OR "both datasets" (redundant)
"the total number of the whole bikes" → Use "the total number of bikes"
"Of course" → Too colloquial for academic writing
"and so on" → Use "etc." is also colloquial; use formal alternatives or list explicitly
In abstract/conclusion: Use "is", "are" (NOT "can be", "may be")
First Principles
The First Principle: "Write Correctly"
Before worrying about style, length, sophistication, or elegance:
Write what you actually did - express your thoughts and methods clearly
Write what you actually want to say - be direct and honest
Get it down first - refinement and polishing come SECOND
Do NOT worry about:
"Is this too long?"
"Is this too simple?"
"Is this too Chinese-style?"
"Is this too wordy?"
Common anxiety that holds writers back:
Fear of imperfect English
Over-thinking sentence structure
Premature polishing before content is complete
Remember:
You can always improve a complete draft
You cannot improve a blank page
Clarity of expression beats stylistic flourishes
Content is king; presentation is queen
When in doubt:
Choose clarity over complexity
Choose consistency over variety
Choose directness over circumlocution
Academic writing prioritizes precision and readability
Summary
This guide provides comprehensive coverage of academic writing pitfalls identified across 100+ papers. Use it as a reference before submission, but remember: the first principle is to write correctly and express your ideas clearly. Refinement comes second.
Key takeaways:
Use active voice and "we" in abstract/contributions
Eliminate Chinese punctuation and colloquialisms
Maintain consistency in terminology, formatting, and style
Focus on clarity over complexity
Write for the reader, not for yourself
Cite authoritative sources (peer-reviewed, not encyclopedias)
Follow LaTeX/Word best practices for academic templates