ZUNA1.1：更靈活的 EEG 去噪與超解像基礎模型
ModelZUNA1.1: A More Flexible EEG Foundation Model for Denoising and Super-ResolutionZyphra releases ZUNA1.1, an Apache 2.0-licensed EEG foundation model that reconstructs, denoises, and upsamples data across arbitrary channel layouts. Designed for real-world recordings, we build upon our previous EEG foundation model, ZUNA1, while matching or exceeding its reconstruction quality and substantially expanding its range and flexibility. ZUNA1.1 now accepts variable length inputs from 0.5-30 seconds and performs reconstructions across arbitrary subsets of channels and time.This post is an update to Zyphra’s ZUNA1.1 release July 16, 2026. The ZUNA1.1 technical paper is now available on arXiv and an updated version of the EEG Playground is available on Zyphra Cloud, now with more features.August 3, 2026
Zyphra 發佈採 Apache 2.0 授權的 EEG 基礎模型 ZUNA1.1，支援在任意通道配置下重建、去噪及升採樣 EEG。模型維持 380M 參數，支援 0.5–30 秒輸入，重建品質達到或超越 ZUNA1。用戶可在 Zyphra Cloud 的 EEG Playground 透過瀏覽器使用，亦可下載模型在本地執行。
來源：Zyphra Research · zyphra.com