yzxoi/paperly-codex/gap-detection

Gap Detection

Overview

Analyzes a team's literature base to identify what has NOT been studied, contested, or resolved. Surfaces research gaps, contradictions left unaddressed, and methodological blind spots — helping teams identify novel contributions before writing.

When to Use

• User asks "what hasn't been done yet in this area?"
• User is preparing a manuscript introduction or contribution framing section
• User asks "where are the open questions in [topic]?"
• Lab wants to avoid duplicating work already done elsewhere
• User wants to position a new project in the existing literature landscape

Key Capabilities

• Scan a paper corpus for explicitly stated limitations and future work sections
• Cluster unstudied conditions, populations, datasets, or modalities
• Cross-reference claimed gaps against actual recent work (via arXiv/PubMed)
• Identify methodological monocultures (everyone uses the same dataset/approach)
• Generate a structured "gap map" organized by type: empirical, theoretical, applied
• Suggest 3–5 concrete research directions with supporting evidence

Usage Examples

Find gaps in a topic area

gaps = gap_detection.analyze(
    topic="self-supervised learning for genomics",
    corpus=review.get_papers(),
    include_future_work_sections=True
)

Cross-check a proposed contribution against the literature

gap_detection.validate_novelty(
    claim="We are the first to apply contrastive learning to single-cell RNA-seq",
    search_depth="comprehensive"
)

Generate gap map for manuscript framing

gap_detection.gap_map(
    topic="federated learning in clinical settings",
    format="structured_markdown",
    audience="paper_reviewer"
)

Output Format

Returns categorized gap list with: gap description, supporting evidence (paper excerpts), estimated research effort, and a novelty confidence score. Optionally formatted as a gap map table.

Best Paired With

• claim_auditor for novelty-risk checks
• related_work_scout when positioning the contribution against nearby prior work

Notes

• Explicitly stated limitations in papers are the highest-signal source of gaps
• Combine with contradiction-detection to find gaps created by conflicting findings
• Novelty validation is probabilistic — always verify manually before submission claims