Auto-Research Systems Lack Scientific Closure

A new paper argues that workflow closure in auto-research systems does not equate to scientific closure. While these systems can complete research-like loops—from idea generation to experiment execution, writing, and self-evaluation—the outputs lack scientific standing. The authors propose that trustworthy auto-research should aim for autonomous execution under non-autonomous epistemic control, not autonomous self-sufficiency. Based on a survey of over 100 recent papers and repositories, and a structured audit of 21 representative systems, the paper identifies three recurring failure patterns: objective collapse (single-proxy targets replacing multi-objective aims), validation collapse (internal self-evaluation replacing independent validation), and acceptance collapse. The findings highlight fundamental limitations in current auto-research approaches.