A Three-Phase Factual Recall Circuit in Gemma-2B and Gemma-12B-IT
中文摘要
研究揭示了 Gemma-2B 与 12B-IT 的三阶段事实检索电路，分析了事实的存储、路由和读取过程，并强调了残差流的核心作用。
English Summary
Activation patching reveals a three-phase factual recall circuit in Gemma-2B and 12B-IT, detailing how facts are stored, routed, and read via the residual stream.
原文节选
Activation patching reveals how facts are stored, routed, and read out across transformer layers, and why the residual stream does most of the work The post A Three-Phase Factual Recall Circuit in Gemma-2B and Gemma-12B-IT appeared first on Towards Data Science.