Related Work Writing

Related Work Writing

Generate publication-quality Related Work sections with proper citations and thematic organization.

Input

• $0 — Current paper draft or method description
• $1 — Collected literature (BibTeX entries, paper summaries, or literature review notes)

References

• Related work writing prompts and strategies: ~/.claude/skills/related-work-writing/references/related-work-prompts.md

Workflow

Step 1: Analyze the Paper's Contributions

• Read the current paper draft (especially Methods and Introduction)
• Identify the key contributions and novelty claims
• List the technical components that need literature context

Step 2: Organize Literature by Theme

Group related papers into thematic clusters:

• Each cluster should represent a research direction or technique
• Common themes: problem formulation, methodology family, application domain, evaluation approach
• Order themes from most to least relevant to your work

Step 3: Write Each Theme Paragraph

For each thematic group: 1. Topic sentence — Introduce the research direction 2. Describe key works — Summarize 2-5 representative papers 3. Compare and contrast — How does each approach differ from yours? 4. Transition — Connect to the next theme or to your contribution

Step 4: Refine

• Ensure every cited paper has a clear reason for inclusion
• Check that your work's novelty is clear from the comparisons
• Verify all \cite{} keys exist in the .bib file
• Aim for 1-2 pages (single column) or 0.5-1 page (double column)

Rules

• Compare and contrast, don't just describe — "Unlike [X] which assumes..., our method..."
• Organize by theme, not chronologically — Group by research direction
• Cite broadly — Not just the most popular papers; include recent and diverse work
• Be fair — Acknowledge strengths of prior work before stating limitations
• Explain inapplicability — If a method could apply to your setting, explain why you don't compare experimentally, or add it to experiments
• Use present tense for established facts — "Smith et al. propose..." or "This approach uses..."
• End with positioning — The final paragraph should clearly position your work relative to all discussed prior work

Related Skills

• Upstream: literature-search, literature-review, citation-management
• Downstream: paper-writing-section
• See also: survey-generation

Related Work Writing Prompts Reference

Per-Section Tips (from AI-Scientist + AgentLaboratory)

Related Work Section Guidelines

- Academic siblings of our work, i.e. alternative attempts in literature at trying to solve the same problem.
- Goal is to "Compare and contrast" — how does their approach differ in either assumptions or method?
- If their method is applicable to our Problem Setting, I expect a comparison in the experimental section.
- If not, there needs to be a clear statement why a given method is not applicable.
- Note: Just describing what another paper is doing is not enough. We need to compare and contrast.
- Organize by theme, not chronologically.
- Cite broadly — not just the most popular papers.

Writing Strategy

Step 1: Thematic Clustering

Organize cited papers into thematic groups. Common themes:

1. Problem formulation variants — Different ways the same problem has been formulated 2. Methodology families — Groups of methods sharing a core technique (e.g., attention-based, graph-based) 3. Application domains — Where similar techniques have been applied 4. Evaluation approaches — Different benchmarks or evaluation paradigms 5. Theoretical foundations — Relevant theoretical results

Step 2: Per-Theme Paragraph Structure

[Topic sentence introducing the theme]
[Key work 1: what they did + how it relates to ours]
[Key work 2: what they did + contrast with work 1 and ours]
[Key work 3: extension or variation]
[Summary: what's missing / our advantage in this theme]

Step 3: Comparison Patterns

Assumption differences:

Unlike \citet{smith2023} who assume access to labeled data, our method operates in a fully unsupervised setting.

Methodology differences:

While \citet{jones2024} employ a two-stage pipeline, our approach integrates feature extraction and classification in an end-to-end manner, avoiding the error propagation inherent in decoupled approaches.

Scope differences:

\citet{chen2023} address the related problem of X in the context of Y. Our work differs in that we consider the more general setting of Z, which subsumes their formulation as a special case.

Complementary work:

Complementary to our approach, \citet{wang2024} propose a method for X that could potentially be combined with our framework to further improve performance.

LaTeX Patterns

Citation Commands

• \cite{key} — Parenthetical: (Smith et al., 2024)
• \citet{key} — Textual: Smith et al. (2024) — preferred for subject position
• \citep{key} — Same as \cite in most styles
• \citeauthor{key} — Just the name: Smith et al.

Example Paragraph

\paragraph{Attention Mechanisms.}
The seminal work of \citet{vaswani2017attention} introduced the Transformer architecture,
which relies entirely on self-attention mechanisms. Subsequent works have sought to
reduce the quadratic complexity of attention. \citet{kitaev2020reformer} propose locality-sensitive
hashing to approximate attention, while \citet{wang2020linformer} project keys and values
to a lower-dimensional space. Unlike these approaches, which sacrifice exact attention
computation for efficiency, our method maintains exact attention while achieving
sub-quadratic complexity through a novel sparse attention pattern that
exploits the inherent structure of the input data.

Refinement Checklist

• [ ] Every cited paper has a clear reason for inclusion
• [ ] The section is organized by theme, not chronologically
• [ ] Each paragraph compares and contrasts, not just describes
• [ ] Our work's novelty is clear from the comparisons
• [ ] All \cite{} / \citet{} keys exist in the .bib file
• [ ] Recent work (last 2-3 years) is well-represented
• [ ] Foundational/seminal papers are included where relevant
• [ ] No self-citations that violate anonymization
• [ ] The section positions our work clearly at the end