JevGPT: a dual-system architecture for autonomous research
A paper submitted to arXiv on 13 October 2026 introduces JevGPT, a dual-system architecture for autonomous research that couples fast evidence processing with deliberative research planning. System Two sets evidence objectives and integrates findings, while System One locates supporting passages, assesses relevance and guides further inspection, with independent source judgments run in parallel. The authors report that on fixed-corpus question answering and document ranking JevGPT achieved leading task performance with faster completion and lower cost than comparably effective research agents; these results come from the paper's own experiments and have not been independently verified.
It suggests that improving how thoroughly and efficiently research agents inspect sources could be a practical route to cheaper, more capable autonomous research.