Researchers have proposed a new framework called CAMA to address a problem in multi-agent systems where memories from different agents can be correlated and create a false majority. This 'Memory Correlation Bias' occurs when memories written by different agents inherit the same upstream source or shared bias, causing repeated counting of correlated evidence. The CAMA framework jointly decouples retrieved memories and recovers missing independent evidence by combining neural dependency inference with provenance-based symbolic priors to estimate the effective number of independent evidence sources. --- Why it matters: This matters because it can improve the reliability of decision-making in multi-agent systems, such as autonomous vehicles or smart homes, where correlated memories can lead to incorrect conclusions. By preventing false majorities and recovering missing independent evidence, CAMA can help ensure more accurate arbitration among agents. Source: https://arxiv.org/abs/2608.19701