gather-citations

name

gather-citations

description

Search for and compile academic citations for a research paper into a BibTeX bibliography. Use this skill when the user wants to find references, build a references.bib file, collect citations for a paper, search for related work, or create a bibliography. Also use when the user mentions "find papers," "related work," "references," or "BibTeX." This skill uses web search and the Semantic Scholar API.

Gather Citations

You are building the bibliography for a scientific paper. Your job is to systematically search for relevant academic papers and compile them into a references.bib file with proper BibTeX entries.

Prerequisites

• idea.md — the research idea (to know what to cite)
• Experiment summaries (baseline, research, ablation JSONs) — to know what methods/datasets were used
• Web search capability (for finding papers)
• SLR/ directory (optional) — systematic literature review markdown files with paper references

Step-by-Step Process

0. Extract Citations from SLR (if available)

If an SLR/ directory exists in the project root, read all .md files in it. Extract any paper references found (titles, authors, years, venues, arXiv IDs, DOIs). For each reference:

1. Create a BibTeX entry and add it to references.bib
2. Use the Semantic Scholar API to verify and complete missing fields (authors, venue, year)
3. Note which citation category each paper covers

This gives you a head start — many needed citations may already be in the SLR.

1. Identify Citation Needs

Read idea.md, the experiment summaries, and slr_synthesis.md (if it exists). Make a checklist of papers you need to find. Mark categories already covered by SLR-extracted citations:

Category checklist:

• The core problem being addressed (seminal papers in the area)
• The specific method/technique proposed (prior work on similar approaches)
• Datasets used (original dataset papers)
• Models and architectures referenced (e.g., ResNet, Transformer, GPT)
• Optimizers and training techniques (Adam, learning rate schedules)
• Evaluation metrics (if using non-standard metrics, cite the source)
• Baseline methods (papers for each baseline compared against)
• Related/competing approaches (recent work on the same problem)
• Theoretical foundations (if the method builds on specific theory)

2. Search for Papers

For each citation need, search using the Semantic Scholar API or web search.

Using Semantic Scholar API (preferred):

curl -s "https://api.semanticscholar.org/graph/v1/paper/search?query=attention+is+all+you+need&limit=5&fields=title,authors,year,venue,externalIds,citationCount,abstract" | python -m json.tool

Using web search:

Search for "{paper title}" site:arxiv.org or "{paper title}" site:semanticscholar.org.

3. Build BibTeX Entries

For each paper found, create a proper BibTeX entry. Use this format:

@article{vaswani2017attention,
  title={Attention is All You Need},
  author={Vaswani, Ashish and Shazeer, Noam and Parmar, Niki and Uszkoreit, Jakob and Jones, Llion and Gomez, Aidan N and Kaiser, {\L}ukasz and Polosukhin, Illia},
  journal={Advances in Neural Information Processing Systems},
  volume={30},
  year={2017}
}

BibTeX key convention: {first_author_lastname}{year}{first_word_of_title} in lowercase. Example: vaswani2017attention.

Entry types:

• @article — journal papers, arXiv preprints
• @inproceedings — conference papers (NeurIPS, ICML, ICLR, etc.)
• @book — textbooks
• @misc — software packages, websites, datasets

4. Iterative Gathering

Work through citations in rounds. Each round:

1. Check the checklist — what's still missing?
2. Search for the most important missing citation
3. Verify the paper exists and is relevant (read the abstract)
4. Create the BibTeX entry
5. Append to references.bib
6. Update the checklist

Deduplication: Before adding a new entry, check if a paper with the same title (case-insensitive) already exists in the file.

Stop when:

• All checklist categories have at least one citation
• You have 15-30 references total (typical for a 4-8 page paper)
• No more important citations are missing

5. Quality Checks

After gathering, verify:

• Every BibTeX key is unique
• No duplicate papers (same paper under different keys)
• Author names are properly formatted ({Last}, First)
• Years are correct
• Venue names are consistent (e.g., always "NeurIPS" not sometimes "NIPS")
• All entries have required fields (title, author, year)
• ArXiv papers have the eprint field

6. Create Citation Annotations

Optionally, create a citation_notes.json that maps each citation to where it should be used in the paper:

{
  "vaswani2017attention": {
    "description": "Introduced the Transformer architecture with self-attention",
    "use_in_paper": "Related work section, when discussing attention mechanisms"
  },
  "he2016deep": {
    "description": "ResNet - deep residual learning",
    "use_in_paper": "Method section, as the backbone architecture"
  }
}

This helps the paper-writing stage place citations correctly.

Common Citation Sources

Source	How to Search	Best For
Semantic Scholar API	curl as shown above	Structured search, getting BibTeX
Google Scholar	Web search with site:scholar.google.com	Broad coverage
arXiv	Web search with site:arxiv.org	Preprints, recent papers
DBLP	https://dblp.org/search?q={query}	Computer science venues

BibTeX Template Reference

% Conference paper
@inproceedings{author2024title,
  title={Paper Title},
  author={Last, First and Last2, First2},
  booktitle={Proceedings of the International Conference on Machine Learning},
  year={2024}
}

% Journal paper
@article{author2024title,
  title={Paper Title},
  author={Last, First and Last2, First2},
  journal={Journal Name},
  volume={42},
  number={1},
  pages={1--15},
  year={2024}
}

% ArXiv preprint
@article{author2024title,
  title={Paper Title},
  author={Last, First and Last2, First2},
  journal={arXiv preprint arXiv:2401.12345},
  year={2024}
}

Output

experiments/{idea_name}/
├── references.bib         # Complete BibTeX bibliography
└── citation_notes.json    # Where each citation should be used (optional)