As large language model-based agents enter the field of autonomous scientific research, their ability to resist pseudoscience becomes increasingly important. Otherwise, these systems may rapidly generate plausible yet misleading studies that contaminate academic literature and erode public trust in science. We present PseudoBench, an adversarial benchmark for evaluating the ability of agentic auto-research systems to identify and resist pseudoscientific narratives.
PseudoBench contains 200 curated pseudoscientific claim-evidence pairs across five domains and evaluates agents through an end-to-end research pipeline from experiments to writing. Testing seven state-of-the-art agents, we find that current systems readily produce persuasive reports that align with pseudoscientific premises, with near-zero refusal rates and the highest resistance of only 27.4%. Stronger agents risk packaging pseudoscience in more sophisticated scientific language, increasing its apparent credibility. These findings reveal an alarming capacity to fuel pseudoscience, calling for scientific alignment before widespread deployment.
Blogger's Review: The PseudoBench study highlights the potential risks of large language models in scientific domains, particularly their ability to propagate pseudoscience. As these technologies advance, ensuring their scientific integrity and accuracy becomes crucial, necessitating enhanced scrutiny and control over their outputs before practical application.