PrediMem

PrediMem is a dual-system robotic memory framework that pairs a high-level vision-language model planner with a low-level vision-language-action executor, using predictive coding to improve keyframe selection and memory-augmented task continuation in dynamic environments. It is a research model artifact, not a commercial product. Source: supabase.