Bringing analytic rigor to agentic AI for science: The Brain Researcher platform for neuroimaging data analysis

TL;DR - Brain Researcher is an agentic platform that embeds methodological constraints, evidence tracking, and scientific review into neuroimaging analysis workflows. It aims to make agent-generated findings more defensible by testing alternative analyses and limiting claims to what the evidence supports.

• Improved first-choice tool-selection accuracy across seven models from 23.3% to 93.6%.
• Increased verifiable grounding from 4.6% to 22.0%.
• Uses multiverse analyses to reveal how findings vary with analytic choices.
• Links decisions to evidence and provenance, while classifying claims as accepted, qualified, revised, blocked, rejected, or deferred.