Causal Memory Policy: Making Memory Utility Identifiable by Intervening on Retrieval

Excerpt

arXiv:2610.02070v1 Announce Type: new Abstract: Memory-augmented large language models must decide which memories to retain, and recent systems do so by estimating each memory's effect on task performance. However, these estimates rely entirely on retrieved memories. When a memory is never retrieved, store-level interventions produce identical outcomes, leaving its utility unidentified. This is a retrieval-level positivity violation, invisible to diagnostics that examine only memory operations.