ProMemAgent is the first proactive agent that incorporates prospective memory into everyday sensory contexts. It organizes past perceptions such as egocentric video and audio into a context‑aware memory, generates prospective cues, and links early events with later developments for future retrieval. When the current context arises, the system retrieves relevant past events and, using memory‑augmented proactive reasoning, decides whether assistance is required.
To evaluate this capability, the authors introduce CPMBench, the first benchmark for long‑horizon proactive assistance over sensory contexts, comprising 1,000 instances across eight everyday categories. Extensive experiments show that ProMemAgent achieves superior memory retrieval performance and consistently outperforms all baselines on long‑horizon proactive assistance tasks.
Review