yzxoi/paperly-codex/cross-paper-synthesis

Cross-Paper Synthesis

Overview

Synthesizes findings across multiple papers into a coherent narrative, structured comparison, or integrated model. Goes beyond summarizing individual papers to produce genuinely novel synthesis — identifying patterns, progressions, and insights that only emerge when reading across the corpus as a whole.

When to Use

• User asks "what's the big picture across all these papers?"
• User wants a related work section written that actually synthesizes, not just lists
• User asks "how has the field's understanding of X evolved over time?"
• User wants to compare methodological approaches across papers
• Lab wants a structured comparison table of models/methods/results

Key Capabilities

• Synthesize across 5–100 papers on a shared topic
• Generate narrative synthesis: what we knew, what changed, what it means
• Generate structured comparison: methods, datasets, metrics, results in table form
• Temporal synthesis: show how understanding has evolved year by year
• Cross-domain synthesis: find connections between papers from different fields
• Identify the "through line" — the core insight that unifies disparate work

Usage Examples

Narrative synthesis of a topic

synthesis = cross_paper_synthesis.narrative(
    papers=review.get_papers(tag="attention-mechanisms"),
    focus="methodological evolution",
    max_words=800,
    audience="expert"
)

Structured comparison table

cross_paper_synthesis.comparison_table(
    papers=review.get_papers(tag="protein-lm-benchmarks"),
    dimensions=["model_size", "training_data", "benchmark", "metric", "score"],
    format="markdown_table"
)

Temporal evolution synthesis

cross_paper_synthesis.timeline(
    topic="in-context learning capabilities",
    papers=review.get_papers(),
    group_by="year"
)

Output Format

Narrative synthesis is prose with inline citations. Comparison tables are Markdown or LaTeX. Timeline output shows year-by-year progression with key papers anchoring each phase.

Best Paired With

• related_work_scout for related-work drafting and literature synthesis
• paper_editor when rewriting a weak compare-and-contrast section

Notes

• Best results with 10–50 papers; >100 papers may require topic narrowing
• Use this skill to distinguish what seems established versus still speculative across the selected corpus
• Always review AI-generated synthesis before including in manuscripts — factual errors are possible