BadScientist: Can a Research Agent Write Convincing but Unsound Papers that Fool LLM Reviewers?

The article presents "BadScientist," a framework designed to assess the ability of AI-generated research papers to deceive LLM-based review systems. It highlights vulnerabilities in automated peer review processes, where fabricated papers can achieve high acceptance rates despite flagged integrity issues, and demonstrates that current integrity checks are largely ineffective. This underscores the necessity for enhanced safeguards in AI-driven scientific publishing to prevent the proliferation of unsound research.