Researchers propose CABLE (Complementary Antecedent-Based Linking and Expansion), a plug-in augmentation for memory retrieval systems. CABLE generates links between memories that complement the direct semantic reach of the host retriever, rather than duplicating it. This allows for better recall of relevant evidence in long-term conversational memory, especially when memories are semantically distant from each other. The authors evaluate CABLE on several benchmark datasets and show that it outperforms existing systems in retrieving supporting evidence. --- Why it matters: This matters to AI researchers because current memory retrieval systems often struggle to recover relevant information from earlier conversations or sessions. CABLE's approach could improve the performance of language models in tasks such as open-domain question-answering, multi-session dialogue, and preference-oriented questions. Source: https://arxiv.org/abs/2608.17911