SeekBrain: An Autonomous Multi-Agent System for Accelerating Neuroscience Discovery

TL;DR - SeekBrain is an autonomous multi-agent framework that uses domain-grounded hierarchical planning to generate hypotheses and analysis pipelines for heterogeneous, multi-scale neuroscience data. It matters because fragmented workflows and data heterogeneity are a key bottleneck in neuroscience, and the system shows agents can produce real scientific findings, not just benchmark scores.

• Builds a repertoire of reusable "analysis recipes" mined from code-paper pairs, coupling codified domain expertise with agentic planning and execution engines.
• Evaluated on BrainArena, an expert-annotated benchmark, where it reportedly outperforms state-of-the-art agent baselines across a range of analysis tasks.
• Deployed in real research: integrated behavioral, neural, and anatomical data to reveal structured distributed neural representations of larval zebrafish behavior, plus a shared axis of regional decoding strength in a mouse decision-making task.
• Emphasis on cross-modal, multi-scale data integration and on-demand pipeline generation rather than fixed, hand-built analysis workflows.