只解构特定输入的神经网络机制，节省算力且精准可控。
Targeted Recovery of Weight-Space Mechanisms From Neural Networks
- 通过高秩通配组件聚焦特定输入，仅分解相关计算路径。
- 在4层模型上用7%算力复现了纯CSS子模型，12层模型可精准重连记忆序列。
- 适合需解释性、低干扰修改模型行为的研究者使用。
参数分解（PD）能将神经网络拆解为可解释的计算单元，忠实反映原模型运作。但扩展至大模型时需巨大算力，成本高且风险大。本文提出目标式参数分解（tPD），通过引入一个高秩通配组件处理非目标数据，仅识别与特定输入（如孤立提示或复杂子任务）相关的组件。我们在玩具模型及基于The Pile训练的Transformer语言模型上验证了tPD，成功恢复出可复现、机制忠实的计算回路。在4块变压器中，仅用其公开分解所需7%的浮点运算量提取出纯CSS子模型；在12块变压器中，对记忆序列实现手术式删除与重连，对其他输入影响极小。
原文摘要 · Abstract (English)
Parameter decomposition (PD) decomposes neural networks into interpretable computational components that faithfully reflect the original network's operations. However, scaling PD to large models requires vast compute, making it a costly and risky endeavor. Here we propose targeted PD (tPD), which identifies only the components that process specific inputs of interest -- from isolated prompts to large subtasks -- by introducing a high-rank catch-all component that handles all non-target data. We validate tPD on toy models and on transformer language models trained on The Pile, where it recovers reproducible, mechanistically faithful circuits. We extract a CSS-only submodel of a 4-block transformer using 7% of the FLOPs of its published decomposition, and in a 12-block transformer we surgically ablate and rewire memorized sequences, with negligible side effects on other inputs.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。