Breaking01 / AI
JevGPT links evidence review with research planning
The dual-system design connects evidence processing to research planning, but the abstract gives no performance scores or costs for individual comparators.

JevGPT researchers presented a dual-system architecture for autonomous research on limited budgets in an abstract dated 10 Nov.
According to the abstract, System Two sets evidence objectives and integrates findings. System One finds supporting evidence, assesses relevance and guides further review.
The design processes judgments about individual sources independently and in parallel, the researchers wrote in the abstract. Selected passages from the original documents inform the next research step.
The researchers reported testing JevGPT on question answering within a fixed collection of documents and on document ranking. Their published results described faster completion and lower costs than research agents with comparable task performance.
The abstract identifies a gap between retrieving relevant sources and using their evidence in research. The researchers reported that JevGPT inspected more sources and covered more documents supporting its answers within the research process.
The researchers did not disclose performance scores or costs for individual comparators in the abstract.