A routing policy that weighs predicted answer quality against wireless and compute cost is claimed to cut response latency by 5-15% and large-model invocations by 10-20% without losing benchmark accuracy.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.IT 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Dynamic Quality-Latency Aware Routing for LLM Inference in Wireless Edge-Device Networks
A routing policy that weighs predicted answer quality against wireless and compute cost is claimed to cut response latency by 5-15% and large-model invocations by 10-20% without losing benchmark accuracy.