Google DeepMind 推出 Gemma Scope 2：支援全新 Gemma 3 家族，深化 AI 安全與可解釋性研究
Original: Gemma Scope 2: helping the AI safety community deepen understanding of complex language model behavior
Google DeepMind has officially released Gemma Scope 2, extending its powerful open-source model interpretability tools to the latest Gemma…
Google DeepMind 宣布推出 Gemma Scope 2，這是一套專為全新 Gemma 3 模型家族設計的稀疏自編碼器（SAE）工具包。Gemma Scope 2 旨在提供強大的「可解釋性」工具，讓研究人員能夠深入剖析模型內部的神經元激活狀態與決策邏輯。此舉將大幅降低 AI 安全與機制可解釋性（Mechanistic Interpretability）研究的門檻，協助社群理解複雜語言模型的行為。
Google DeepMind has officially released Gemma Scope 2, extending its powerful open-source model interpretability tools to the latest Gemma 3 model family. This release is intended to help the AI safety community and academia gain deeper insight into and analysis of the internal workings of complex large language models (LLMs).
Free shows the 3-line summary; Pro unlocks the full deep summary (~300 words) so you never have to click through.
See Pro plans →Want the original English / full article?
Read on Google DeepMind Blog →Summaries are AI-generated; the original article is authoritative.