An arXiv abstract for DataPraxis (arXiv:2610.26924v1) opens by positioning professional analytics as a demanding setting for general-purpose agents. It argues that such work requires sustained reasoning, iterative investigation, and reliable evidence.
The paper's title identifies the proposed mechanism: scaling analytics tasks via verified workspace evolution. That suggests agents are meant to work in a workspace that changes over time, with verification used to keep the process trustworthy. The abstract starts to say that training these capabilities requires diverse and ... but the announcement text is truncated, so the full methodology is not available from the abstract alone.
This summary is based solely on the arXiv abstract; no other source was provided for comparison.