Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GAPMAP: Mapping Scientific Knowledge Gaps in Biomedical Literature Using Large Language Models

Nourah M Salem, Elizabeth White, Michael Bada, Lawrence Hunter

TL;DR

This work addresses the automatic discovery of knowledge gaps in biomedical literature by distinguishing explicit gaps signaled in text from implicit gaps inferred from discourse. It introduces the TABI framework for structured implicit-gap reasoning and conducts two experiments across four datasets to benchmark open- and closed-weight LLMs on paragraph- and full-document inputs. The study provides datasets and evaluation methods for explicit and implicit gaps, including IPBES and COVID-19 explicit-gap benchmarks and two implicit-gap datasets (paragraph-level and full-text). Findings show that large LLMs can robustly identify both gap types, with prompting strategy and input scale significantly influencing performance, and highlight practical deployment considerations for gap recommender systems. The results support using heterogeneous LLMs and human-in-the-loop verification to guide early-stage research and science-policy decision-making.

Abstract

Scientific progress is driven by the deliberate articulation of what remains unknown. This study investigates the ability of large language models (LLMs) to identify research knowledge gaps in the biomedical literature. We define two categories of knowledge gaps: explicit gaps, clear declarations of missing knowledge; and implicit gaps, context-inferred missing knowledge. While prior work has focused mainly on explicit gap detection, we extend this line of research by addressing the novel task of inferring implicit gaps. We conducted two experiments on almost 1500 documents across four datasets, including a manually annotated corpus of biomedical articles. We benchmarked both closed-weight models (from OpenAI) and open-weight models (Llama and Gemma 2) under paragraph-level and full-paper settings. To address the reasoning of implicit gaps inference, we introduce \textbf{\small TABI}, a Toulmin-Abductive Bucketed Inference scheme that structures reasoning and buckets inferred conclusion candidates for validation. Our results highlight the robust capability of LLMs in identifying both explicit and implicit knowledge gaps. This is true for both open- and closed-weight models, with larger variants often performing better. This suggests a strong ability of LLMs for systematically identifying candidate knowledge gaps, which can support early-stage research formulation, policymakers, and funding decisions. We also report observed failure modes and outline directions for robust deployment, including domain adaptation, human-in-the-loop verification, and benchmarking across open- and closed-weight models.

GAPMAP: Mapping Scientific Knowledge Gaps in Biomedical Literature Using Large Language Models

TL;DR

This work addresses the automatic discovery of knowledge gaps in biomedical literature by distinguishing explicit gaps signaled in text from implicit gaps inferred from discourse. It introduces the TABI framework for structured implicit-gap reasoning and conducts two experiments across four datasets to benchmark open- and closed-weight LLMs on paragraph- and full-document inputs. The study provides datasets and evaluation methods for explicit and implicit gaps, including IPBES and COVID-19 explicit-gap benchmarks and two implicit-gap datasets (paragraph-level and full-text). Findings show that large LLMs can robustly identify both gap types, with prompting strategy and input scale significantly influencing performance, and highlight practical deployment considerations for gap recommender systems. The results support using heterogeneous LLMs and human-in-the-loop verification to guide early-stage research and science-policy decision-making.

Abstract

Scientific progress is driven by the deliberate articulation of what remains unknown. This study investigates the ability of large language models (LLMs) to identify research knowledge gaps in the biomedical literature. We define two categories of knowledge gaps: explicit gaps, clear declarations of missing knowledge; and implicit gaps, context-inferred missing knowledge. While prior work has focused mainly on explicit gap detection, we extend this line of research by addressing the novel task of inferring implicit gaps. We conducted two experiments on almost 1500 documents across four datasets, including a manually annotated corpus of biomedical articles. We benchmarked both closed-weight models (from OpenAI) and open-weight models (Llama and Gemma 2) under paragraph-level and full-paper settings. To address the reasoning of implicit gaps inference, we introduce \textbf{\small TABI}, a Toulmin-Abductive Bucketed Inference scheme that structures reasoning and buckets inferred conclusion candidates for validation. Our results highlight the robust capability of LLMs in identifying both explicit and implicit knowledge gaps. This is true for both open- and closed-weight models, with larger variants often performing better. This suggests a strong ability of LLMs for systematically identifying candidate knowledge gaps, which can support early-stage research formulation, policymakers, and funding decisions. We also report observed failure modes and outline directions for robust deployment, including domain adaptation, human-in-the-loop verification, and benchmarking across open- and closed-weight models.

Paper Structure

This paper contains 24 sections, 3 equations, 2 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Our main contributions are as follows:
  3. Related Work
  4. Datasets and Models
  5. Explicit Knowledge Gaps
  6. IPBES ipbes_knowledge_gaps_page5
  7. Scientific Challenges & Directions lahav2022search
  8. Implicit Knowledge Gaps
  9. Paragraph-Level Manual Implicit-Gap Dataset
  10. Full-Text Comprehension Pilot Dataset
  11. Models
  12. Experimental Setup
  13. Explicit Knowledge Gaps
  14. Chunking Strategy
  15. Implicit Knowledge Gap Identification
  16. ...and 9 more sections

Figures (2)

  • Figure 1: Model agreement and complementarity.Left: For each model, predictions unique to that model (orange) vs. those shared with $\geq$1 other model. Right: Overlap among the top four models. Numbers denote instance counts per region.
  • Figure 2: Model performance across knowledge-gap categories (Research Aim, Levels of Evidence, Anomaly/Curious Findings, Barriers, and Future Opportunities). The chart shows relative strengths of GPT-5, GPT-4o Mini, Llama-3.3-70B, and Llama-4-17B normalized.