interpretability

Qwen-Scope sparse autoencoders visualization showing neural network feature decomposition with glowing activation nodes

Alibaba Qwen Releases Qwen-Scope, Open Suite of Sparse Autoencoders

Alibaba releases Qwen-Scope, an open suite of sparse autoencoders enabling direct manipulation of model features, data classification, and model behavior analysis without prompt engineering.

Shima Zamil · 1 May 2026