Abstract Summary

This workshop version extends the Mugi idea to nonlinear LLM operations and frames value-level parallelism as a broader execution model.

Research Context

This paper contributes to my research program in Mugi, workshop paper, WUC 2026. It is part of the broader work on efficient ML systems, hardware-software co-design, and deployment-aware computer architecture.

Mugiworkshop paperWUC 2026

Related Papers