AI can now write research papers — researchers built the tool and the test to prove it works

Researchers created an AI system that takes raw experimental data and notes, then produces a finished research paper with literature review and diagrams. This is the first time someone built both the system and a standardized test (using real papers from top conferences) to measure whether the output is actually good.

Until now, AI paper-writing systems were one-off tools bolted onto specific experiments. This one is modular — it takes whatever materials you have and produces submission-ready output. The benchmark matters more than the system itself: for the first time, you can measure whether an AI-written paper would actually pass peer review or just look plausible on the surface. Human raters preferred PaperOrchestra's literature reviews to those from existing systems by 50-68 percentage points, and preferred the full manuscripts by 14-38 percentage points. The real question is whether this becomes a deployment tool (researchers actually using it) or stays a research proof-of-concept (the system works, but nobody ships it).