Paging the Experts: A Reproducible Characterization of Flash-Backed MoE Inference on iPhone

arXiv:2609.29032 · cs.PF, cs.LG · Submitted 2026-09-24 · Read on arXiv

cs.PF, cs.LG

Submitted: 2026-09-24

Updated: 2026-09-24

Terminology

• Trace — 8× in this paper · explained in 2 episodes
• Decode — 5× in this paper · explained in 2 episodes
• Allocator — 3× in this paper · explained in 1 episode
• Benchmark — 3× in this paper · explained in 2 episodes
• Quantization — 3× in this paper · explained in 12 episodes
• Accuracy — 2× in this paper · explained in 1 episode
• Calibration — 2× in this paper · explained in 7 episodes
• Sparsity — 2× in this paper · explained in 4 episodes
• Activation Function — 1× in this paper · explained in 1 episode
• Activation Sparsity — 1× in this paper · explained in 1 episode
• Alignment — 1× in this paper · explained in 1 episode
• Provenance — 1× in this paper · explained in 4 episodes

Sources

• LLM in a flash: Efficient Large Language Model Inference with Limited Memory
• PowerInfer-2: Fast Large Language Model Inference on a Smartphone
• Fiddler: CPU-GPU Orchestration for Fast Inference of Mixture-of-Experts Models

Related papers

• A Controlled Study of Memory Hierarchy Transitions in Quantum Circuit Simulation on Apple M4 Pro Unified Memory Architecture
• The Caf'e in Amsterdam: When the Incumbent Becomes the Oracle