More Agents, More Reliable? System-Level Reliability of Multi-Agent LLMs under Operational-Profile Uncertainty
4.40T1 sourcearXiv cs.MA
Source record
Published by arXiv cs.MA (T1 source). The original is at https://arxiv.org/abs/2611.07140.
Pipeline notes
The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.
SummaryThe paper extends HIP-LLM with a Bayesian network that propagates uncertainty in agent failures and answer-merging rules under changing operational inputs. In a chemical-hazard extraction pipeline, three-model majority voting did not clearly outperform the best model, and voter failures could remove its advantage.
Why it mattersIt provides a reusable reliability assessment that treats changing demand mixes and aggregation failures as design variables rather than reporting benchmark averages alone.
Cited by
No citations on record.