Zyphra Releases ZUNA1.1: An Apache 2.0 EEG Foundation Model With Variable-Length Inputs From 0.5 To 30 Seconds

Short summary

Zyphra released ZUNA1.1, a 380M-parameter Apache 2.0-licensed foundation model for scalp-EEG data that reconstructs, denoises, and upsamples signals across arbitrary channel layouts. The model accepts variable-length inputs from 0.5 to 30 seconds, a significant upgrade over ZUNA1's fixed five-second window, while maintaining or improving reported NMSE. The release is open-source under Apache 2.0, making it accessible for neuroscience and healthcare signal-processing research.

• •ZUNA1.1 is a 380M masked diffusion autoencoder for EEG under Apache 2.0
• •Supports variable-length inputs from 0.5 to 30 seconds vs ZUNA1's fixed 5 seconds
• •Maintains or improves NMSE while widening the input range

Generated with AI, which can make mistakes.