Source record

Published by arXiv cs.MA (T1 source). The original is at https://arxiv.org/abs/2609.18283.

Pipeline notes

The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.

Research paper introducing agentic-eCAL, a metric generalizing Energy Cost of AI Lifecycle to multi-agent workflows across edge-cloud tiers. Validated on A100 and H100 GPUs with 16 open-weight models and 8 orchestration topologies. Key finding: inter-agent text transport accounts for only 0.25% of workflow energy, so placement costs are driven by induced inference and context processing, not communication itself.

First empirical energy benchmark for multi-agent execution graphs. Reusable metric plus a counter-intuitive placement finding (comms cost negligible) that should inform edge-cloud agent architecture decisions.

Cited by

No citations on record.