$\Sigma$-Mem: An Online Reliability Memory for LLM-based Multi-Agent Systems
3.40T1 sourcearXiv cs.MA
Source record
Published by arXiv cs.MA (T1 source). The original is at https://arxiv.org/abs/2607.27958.
Pipeline notes
The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.
SummaryResearch paper introducing Σ-Mem, an online reliability memory for LLM multi-agent systems that tracks per-peer competence and peer-relationship evidence via bounded spectral updates. Enables stable adaptation without retraining, applicable to peer routing and weighted voting, outperforming majority voting on OOD evaluation across five Qwen models.
Why it mattersProposes a reusable reliability memory layer for multi-agent LLM systems with bounded spectral updates and demonstrated generalization to unseen peers and tasks.
Cited by
No citations on record.