A Geometric Calculator Inside a Neural Network
Researchers discover a general-purpose addition module in Llama 3.1 8B that uses circular representations (Fourier features) to encode numbers and perform addition, reused across tasks like month and day arithmetic. The module resides in layer 18, computes modular sums in parallel, and causal steering confirms its role.
A Geometric Calculator Inside a Neural Network
Report an error
The correction channel is not available yet. You can copy the article reference below for later.
Correction instructionsSummary only
Under the source content permission, this page provides an editorial summary, key points, and a link to the original article.