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

LiRA: A Multi-Agent Framework for Reliable and Readable Literature Review Generation

arXiv 2025 54 method

TLDR

LiRA is a multi-agent framework that automates literature review writing, improving readability and accuracy via specialized agents.

Reasoning

The paper addresses an under-explored area (writing phase of reviews) with a novel multi-agent approach and includes real-world evaluations, but its scope is limited to literature review generation rather than broader scientific discovery or experimentation.

Read-first score

Read-first score 54, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 57.

Recency 8%
86.7

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

Methodology quality 25%
70

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

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

Reproducibility 25%
38

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

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 37.

Keyword Scores

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

Deep Analysis

Innovations

  • Multi-agent collaborative workflow that emulates the human literature review process
  • Specialized agents for content outlining, subsection writing, editing, and reviewing
  • Focus on improving readability and factual accuracy in automated review generation

Methodology

LiRA is a multi-agent framework with agents for outlining, writing, editing, and reviewing, designed to generate literature reviews. It is evaluated on SciReviewGen and a proprietary ScienceDirect dataset against AutoSurvey and MASS-Survey baselines, using metrics of writing quality, citation quality, and similarity to human reviews, along with real-world retrieval and robustness tests.

Key Results

LiRA outperforms AutoSurvey and MASS-Survey in writing and citation quality while maintaining competitive similarity to human-written reviews, and shows robustness to reviewer model variation.

Tags

CL