LifeSide: Benchmarking Agents as Lifelong Digital Companions

2026-06-03

AI Summary

Current evaluation methods struggle to assess digital companion agents’ ability to integrate cross-session cues, maintain consistent user understanding, and adapt to evolving privacy boundaries over long-term interactions. This work proposes the first multidimensional evaluation framework tailored for lifelong digital companionship, introducing a recurrent benchmark that couples memory, affect, and contextual dynamics across multiple sessions. Leveraging multi-agent simulation, users are modeled as persistent personas with hierarchical profiles and event trajectories. Experiments across 2,000 personas and 111,000 tasks reveal that even models excelling on existing memory benchmarks fail to sustain accurate user modeling and authentic companionship over time, exposing fundamental limitations in current systems’ capacity for long-term interaction.

Abstract

Lifelong digital companions must integrate cross-session cues, continually update their understanding of users, and adapt to shifting privacy boundaries. Existing evaluations fail to capture this, testing memory recall and short-term empathy in isolation. To bridge this gap, we introduce \benchmark, a benchmark centered on multi-session \textit{Memory-Emotion-Environment} loops. By modeling users as persistent worlds with layered profiles and event trajectories, \benchmark uses multi-agent simulation to project environmental dynamics into dialogue, preserving the critical gap between latent thoughts and observable expressions. Evaluating 2,000 personas and 111K tasks across memory tracking, user understanding, privacy control, and emotional companionship, our experiment results reveal a stark reality: even models that saturate current memory benchmarks fail to sustain accurate user understanding and true companionship over long horizons.

Problem

• lifelong digital companions
• cross-session memory
• user understanding
• privacy boundaries
• emotional companionship

Innovation

• lifelong digital companion
• multi-session benchmark
• Memory-Emotion-Environment loop
• multi-agent simulation
• user modeling