Google’s AI paper reviewer handled about 10,000 conference papers
Google built Paper Assistant Tool, an system for checking research papers before formal review. It reads full papers, checks theoretical results, reviews experiments, suggests improvements, and looks for possible flaws.
The tool was tested as a pre-submission helper at STOC and ICML, two major computer science conferences, and handled about 10,000 papers with around a 30-minute turnaround. On a math-error test, it found 34% more errors than .
Its main technique is , which means using more model work than a single quick answer to search for deeper issues. ers still make the final decisions, while the tool is meant to reduce obvious or serious problems before review.
Key points
- Paper Assistant Tool is an system that checks full research papers for flaws and improvements.
- It was tested at STOC and ICML as a helper before official .
- It handled about 10,000 papers and returned results in around 30 minutes.
- It found 34% more math errors than on a benchmark test.
- It supports ers rather than replacing the final review decision.