RESEARCHMonitorWATCHLIST
Delayed Verification Destabilizes Multi-Agent LLM Belief: Instability Thresholds and Optimal Corrector Placement
arXiv cs.CL — Computation and Language
Factual evidence
What the source reports
Researchers modeled delayed verification in multi-agent LLMs, defining stability thresholds where late corrections allow hallucination propagation.