Reproduce “From Voiceprints to User Entities: Online Speaker-Memory Construction for Multi-User Spoken Agents” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yucong Zhang, Cancan Li, Bo Xiao, Ke Zeng, Ming Li
EvoSpeakerMem routes spoken turns into user-specific memory partitions using evolving speaker evidence, rather than treating a voiceprint as a permanent identity. It separates turn assignment from acoustic-state updates, allowing selective updates after a link. On 4,113-query LoCoMo-ID conditions, Gated Update achieves 75.27% clean and 71.77% shifted identity-gated QA with zero wrong-user authorizations, on synthetic voices and simulated shifts.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.