AgentSLR: automating systematic literature reviews in epidemiology with agentic AI

Conference item

Systematic literature reviews are essential for synthesizing scientific evidence but are costly, difficult to scale and time-intensive, creating bottlenecks for evidence-based policy. We study whether large language models can automate the complete systematic review workflow, from article retrieval, article screening, data extraction to report synthesis. Applied to epidemiological reviews of nine WHO-designated priority pathogens and validated against expert-curated labels, our open-source agentic harness (AgentSLR) achieves performance comparable to human researchers while reducing review time from approximately 7 weeks to 20 hours (a 58× speedup). Our comparison of five frontier models reveals that performance on SLR is driven less by model size or inference cost than by each model’s distinctive capabilities. Through human-in-theloop validation, we identify key failure modes. Our results demonstrate that agentic AI can substantially accelerate scientific evidence synthesis in specialised domains.

Authors: Padarha, S; Kearns, RO; Naidoo, T; Yang, L; Borchmann, L

• Publication status: Accepted
• Peer review status: Peer reviewed
• Host title: International Conference on Machine Learning (ICML) 2026 AI for Science Workshop
• Acceptance date: 2026-05-01
• Event title: 43rd International Conference on Machine Learning
• Event location: Seoul, South Korea
• Event website: https://ai4sciencecommunity.github.io/icml26.html
• Event start date: 2026-07-10
• Event end date: 2026-07-10
• Language: English