SpeakerMem-R1 targets long-term memory for multi-party conversational AI, where systems must track who said what, who a statement concerns, how participants perceive each other, and how group state evolves over time — details existing general-purpose LLM memory systems tend to lose. It uses a dual-track memory that stores speaker-labeled verbatim messages alongside derived state organized at both the person level and the group level, targeting two bottlenecks: message attribution/relational understanding, and state reconstruction from interleaved multi-party histories.
