A-IO adaptively orchestrates LLM inference on NPUs to address memory bottlenecks, model scaling paradoxes, and synchronization costs in speculative decoding.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DC 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
A-IO: Adaptive Inference Orchestration for Memory-Bound NPUs
A-IO adaptively orchestrates LLM inference on NPUs to address memory bottlenecks, model scaling paradoxes, and synchronization costs in speculative decoding.