AutoSynthesis: Multi-Agent Automation for Quantitative Evidence Synthesis
AutoSynthesis is an end-to-end multi-agent system that automates major steps of quantitative evidence synthesis, from a natural-language research question to a PRISMA-aligned meta-analysis report.
TL;DR — AutoSynthesis is described as an end-to-end multi-agent system for automated meta-analysis. From a natural-language research question, it formulates a search strategy, retrieves literature, screens studies, assesses full-text eligibility, extracts quantitative statistics, computes standardized effect sizes, performs random-effects meta-analysis, supports heterogeneity and risk-of-bias assessment, and outputs a PRISMA-aligned report. In the reported application, it screened over 28 studies, extracted more than 20 quantitative claims, and produced pooled estimates described as similar to expert-conducted Hedges' g meta-analyses; exact accuracy metrics are not given in the abstract.
What the abstract supports
The abstract frames evidence synthesis as a way to turn primary research into reliable knowledge for science, medicine, education, and policy. It also identifies a bottleneck: quantitative evidence synthesis remains largely manual and difficult to scale.
AutoSynthesis is introduced as a response to that bottleneck. The abstract describes it as an end-to-end multi-agent system for automated meta-analysis. The supported claim is not merely that the system summarizes papers, but that it automates a full quantitative synthesis workflow from question formulation through statistical pooling and reporting.
The abstract does not provide implementation details such as model names, prompts, tools, search backends, evaluation data, code, or provenance format. Those details would be needed to independently assess reproducibility and robustness.
Full analysis, extracted claims, numerical results, entity graph, FAQ, related work, applications, and BibTeX are available via x402 micropayment.