Hugging Face has introduced MobileMem, a benchmark and framework designed to evaluate on-device long-term memory in AI agents. The initiative addresses the growing need for persistent personal assistants that can understand, remember, and continuously learn from user experiences over extended periods.
MobileMem is grounded in a year-scale collection of mobile experiences, employing a knowledge-grounded synthesis pipeline to construct coherent and temporally consistent long-horizon trajectories from user-app sessions. It provides complementary text and multimodal settings that cover multi-hop and temporal reasoning, knowledge updating, and implicit preference inference.
The framework enables agents to remember the past, understand the present, and adapt to the future. By modeling experiences rather than isolated facts, MobileMem moves memory beyond simple information retrieval toward experiential intelligence for continuous personal learning.
This work highlights the limitations of existing benchmarks in realistic mobile settings, where experiences are heterogeneous, multimodal, evolving, and deeply personal. MobileMem aims to fill this gap by providing a more realistic evaluation environment for next-generation AI assistants.