Nonconformal Distortion and Local Orientation Reversal in the Fourier Circuit of a Modular-Addition Transformer
arXiv cs.AIen
arXiv cs.AI
AI Global WirearXiv:2611.00414v1 Announce Type: new Abstract: We study the local differential geometry of a Fourier-based modular-addition circuit in a one-layer transformer whose algorithm has previously been reverse-engineered. For each of the five key Fourier frequencies, we restrict the model's multilayer perceptron (MLP) contribution to a corresponding Fourier input plane and readout, and measure its directional Beltrami coefficient under continuous input perturbations. Whereas nondegenerate restrictions of ideal complex multiplication are conformal, the trained maps exhibit nonconformal distortion (median $|\mu|=0.304$--$0.446$), spatial organization in the Beltrami argument, and regions of local or
This is a short summary published by AI Global Wire. The full article is owned and hosted by arXiv cs.AI — open it there to read it in full.
Read the full story at arXiv cs.AI- Forskning
Related AI news
- South Korea urged to link chip fab support to local suppliersDIGITIMES · November 3, 2026
- Thunder Tiger flexes unmanned surface vessel capabilities alongside US companiesDIGITIMES · November 3, 2026
- OpenAI partners with Cathedral to test its security defenses and revamps its Security Bug Bounty program, including quadrupling rewards for critical findings (OpenAI)Techmeme · November 3, 2026
- AuroraWorld: Structured 4D World Modeling for Exo-to-Ego Video Generation in Robot ManipulationarXiv cs.AI · November 3, 2026
- MemU-Bench: Benchmarking GUI Agents on Long-Horizon User InteractionsarXiv cs.AI · November 3, 2026
- Exo: An Architecture for Self-Improving Systems at RuntimearXiv cs.AI · November 3, 2026