Awesome Auto Research Hub Papers · Datasets · Projects
← Back to papers

Deep Literature Survey Automation with an Iterative Workflow

arXiv 2025 69.8 method

TLDR

Proposes an iterative workflow for automated literature survey generation using recurrent outline generation and paper cards, outperforming baselines.

Reasoning

Strengths include a novel iterative approach that mimics human reading, paper cards for faithful grounding, and a new benchmark Survey-Arena. Weaknesses are that it focuses only on survey generation rather than full scientific discovery, and evaluation may still rely on subjective human judgment.

Read-first score

Read-first score 69.8, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 57.

Methodology quality 25%
90

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

Recency 8%
86.7

Uses a gentle age decay so recent papers surface without erasing older foundations. 2025

Reproducibility 25%
81

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

Topical relevance 42%
47.5

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: 78.

Keyword Scores

survey generation
10
literature review agent
9
research automation
8
automated research
7
AI for scientific research
6
paper writing agent
5
automated scientific discovery
3
autonomous research agent
3
AI scientist
2
scientific discovery agent
2
automated experimentation
1
experiment design agent
1

Deep Analysis

Innovations

  • Iterative workflow with recurrent outline generation where a planning agent incrementally retrieves, reads, and updates the outline
  • Paper cards that distill each paper into contributions, methods, and findings for faithful paper-level grounding
  • Review-and-refine loop with visualization enhancement to improve textual flow and integrate multimodal elements (figures, tables)
  • Survey-Arena, a pairwise benchmark for more reliable assessment of machine-generated surveys relative to human-written ones

Methodology

The framework uses a planning agent that incrementally retrieves papers, reads them, and updates the outline iteratively. Paper cards summarize each paper's contributions, methods, and findings. A review-and-refine loop with visualization enhancement improves text and integrates figures/tables. Evaluation on established and emerging topics against state-of-the-art baselines using metrics of content coverage, structural coherence, and citation quality, plus a new pairwise benchmark Survey-Arena.

Key Results

The proposed method substantially outperforms state-of-the-art baselines in content coverage, structural coherence, and citation quality, producing more accessible and better-organized surveys.

Tags

CLAI