SurveyGraph: Generating Academic Surveys via Methods-Entity Knowledge Graphs

arXiv 2026 70.2 method

TLDR

SurveyGraph generates academic surveys using a Methods-Entity Knowledge Graph, improving retrieval, outline quality, citation support, and semantic relation extraction across evaluations.

Reasoning

The paper introduces a graph-centric survey generation framework with strong empirical validation, including statistical improvements and human adjudication. Weakness is that end-to-end comparisons are limited to concurrent systems rather than matched backend, so holistic system-level gains remain less clear.

Read-first score

Read-first score 70.2, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 45.

Methodology quality 25%

100

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=baseline,benchmark,evaluation,metric,result

Recency 8%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Reproducibility 25%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=True; dataset=False; markers=artifact,code,github

Topical relevance 42%

37.5

Uses existing LLM keyword relevance scores normalized to 0-100. AI scientist,automated scientific discovery,autonomous research agent,automated research,literature review agent,survey generation,automated experimentation,experiment design agent,AI for scientific research,paper writing agent,research automation,scientific discovery agent

Field roles

FrontierMethodology anchorReproducibility anchor

Rank sensitivity

Stability: volatile; rank range: 153.

Keyword Scores

survey generation

literature review agent

research automation

AI for scientific research

automated research

paper writing agent

autonomous research agent

AI scientist

automated scientific discovery

scientific discovery agent

automated experimentation

experiment design agent

Deep Analysis

Innovations

Methods-Entity Knowledge Graph (MEKG) encoding method-level follow-on, competition, and shared-resource relations instead of treating papers as flat documents
SurveyGraph framework using the MEKG to drive core-paper discovery, graph-aware retrieval, hybrid outlining, and evidence-gated narrative synthesis
Semantic evidence gate for citation grounding during generation
Positioning MEKG as an auditable structural layer for typed diagnostics, relation-specific filtering, and controlled survey generation

Methodology

SurveyGraph uses a Methods-Entity Knowledge Graph (MEKG) encoding method-level relations to support core-paper discovery, retrieval, hybrid outlining, and evidence-gated narrative synthesis. Evaluation includes a controlled seven-method equal-budget benchmark, an archived graph-native diagnostic against flat lexical baselines, semantic edge and citation audits, human adjudication, a SurveyBench-facing collection probe, and same-backend Qwen diagnostics. Metrics include recall, AUC, citation support, coverage, and Cohen's kappa.

Key Results

Graph-aware retrieval nearly doubled recall over flat lexical baselines across eight runs (p = 0.004), and a semantic evidence gate achieved AUC = 0.75. Audits found 80.6% anchored-or-partial citation support across 160 generated sentences and 68.5% supported-or-partial semantic edges across 108 human-adjudicated edges, with improves strongest at 80.6% and extends weakest at 52.8%; a SurveyBench-facing probe reached mean reference coverage 0.521 across four topics with an eight-paper budget.

Limitations

Concurrent-system comparisons provide bounded context rather than matched-backend end-to-end evidence.
Semantic edge support was incomplete, with only 68.5% supported-or-partial relations and extends relations weakest at 52.8%.
SurveyBench-facing collection coverage remained moderate at 0.521 mean reference coverage across four topics with an eight-paper budget.

Resources

Paper Code