The abstract of a new arXiv paper (2605.19156) addresses the current state of auto-research systems. It notes that these systems can now produce complete papers, but argues that this capability alone is not evidence of quality.

The authors point out that the field lacks a systematic study of how good agent-generated papers actually are. They say they introduce such a study, though the provided abstract does not detail its methods or findings.

The paper's central distinction is between feasibility and quality. Producing a paper is not the same as producing a good one, and the authors position their work as a step toward measuring that difference.