In a one-review benchmark, GPT-4 outperformed Claude-3 and Mistral on search strategy, screening, and data extraction, while Claude-3 led on PICO design.
On the Opportunities and Risks of Foundation Models for Natural Language Processing in Radiology
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.IR 1years
2025 1verdicts
REJECT 1roles
background 1polarities
support 1representative citing papers
citing papers explorer
-
Large language models streamline automated systematic review: A preliminary study
In a one-review benchmark, GPT-4 outperformed Claude-3 and Mistral on search strategy, screening, and data extraction, while Claude-3 led on PICO design.