Targeted Recovery of Weight-Space Mechanisms From Neural Networks

Researchers introduced targeted parameter decomposition (tPD), a method that efficiently identifies neural network components processing specific inputs, reducing computational costs. Validated on toy models and transformer language models, tPD recovered mechanistically faithful circuits and enabled surgical ablation of memorized sequences with minimal side effects.

arXiv:2607.13047v1 Announce Type: new Abstract: 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.