This research applies Large Language Model technology to neuroscience by implementing Transformer architecture to generate and predict neural activity patterns across different experimental contexts.

• First bidirectional framework that accurately translates between in vitro and in vivo neural spike data
• Introduces Dice loss function specifically optimized for binary neural signals
• Achieves high-precision generation of neural activity patterns that can replicate complex neuronal behaviors

This breakthrough has significant medical implications by creating a computational bridge between lab experiments and clinical studies, potentially reducing animal testing and accelerating neurological disorder research.

Original Paper: In vitro 2 In vivo: Bidirectional and High-Precision Generation of In Vitro and In Vivo Neuronal Spike Data