Can Large Language Models Match the Conclusions of Systematic Reviews?

2025-05-28

AI Summary

This study investigates whether large language models (LLMs) can reliably reproduce conclusions of clinical expert systematic reviews given identical input literature. Method: We introduce MedEvidence, a benchmark comprising 100 high-quality systematic reviews and their underlying primary studies, and rigorously evaluate 24 general-purpose and biomedical LLMs spanning 7B–700B parameters. Using a multi-model comparative framework with standardized prompt engineering, conclusion consistency scoring, and evidence traceability analysis, we systematically assess cross-study evidence integration. Contribution/Results: We uncover fundamental limitations—including pervasive overconfidence, deficient scientific skepticism, and severe performance degradation on long-context inputs—that persist across model scales, architectural advances, and biomedical fine-tuning. Notably, neither enhanced reasoning capabilities, increased parameter count, nor domain-specific adaptation improves conclusion accuracy; in fact, medical fine-tuning degrades performance. These findings challenge prevailing optimization paradigms and call for a new, evidence-informed AI evaluation framework grounded in clinical decision-making rigor.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Systematic reviews (SR), in which experts summarize and analyze evidence across individual studies to provide insights on a specialized topic, are a cornerstone for evidence-based clinical decision-making, research, and policy. Given the exponential growth of scientific articles, there is growing interest in using large language models (LLMs) to automate SR generation. However, the ability of LLMs to critically assess evidence and reason across multiple documents to provide recommendations at the same proficiency as domain experts remains poorly characterized. We therefore ask: Can LLMs match the conclusions of systematic reviews written by clinical experts when given access to the same studies? To explore this question, we present MedEvidence, a benchmark pairing findings from 100 SRs with the studies they are based on. We benchmark 24 LLMs on MedEvidence, including reasoning, non-reasoning, medical specialist, and models across varying sizes (from 7B-700B). Through our systematic evaluation, we find that reasoning does not necessarily improve performance, larger models do not consistently yield greater gains, and knowledge-based fine-tuning degrades accuracy on MedEvidence. Instead, most models exhibit similar behavior: performance tends to degrade as token length increases, their responses show overconfidence, and, contrary to human experts, all models show a lack of scientific skepticism toward low-quality findings. These results suggest that more work is still required before LLMs can reliably match the observations from expert-conducted SRs, even though these systems are already deployed and being used by clinicians. We release our codebase and benchmark to the broader research community to further investigate LLM-based SR systems.

Problem

• Can LLMs match expert conclusions in systematic reviews?
• Evaluating LLM performance on clinical evidence synthesis.
• Assessing LLM limitations in scientific skepticism and accuracy.

Innovation

• Benchmarking 24 LLMs on MedEvidence dataset
• Evaluating reasoning and model size impact
• Analyzing overconfidence and skepticism in LLMs