Proactive Memory for Ad-Hoc Recall over Streaming Dialogues

TL;DR

This paper introduces ProStream, a proactive memory framework for streaming dialogues that balances reasoning fidelity and latency, supported by a new benchmark called STEM-Bench.

Contribution

It presents ProStream, a hierarchical, adaptive memory system enabling ad-hoc recall in streaming dialogues, and introduces STEM-Bench for evaluating memory in infinite-horizon conversations.

Findings

ProStream achieves higher reasoning fidelity than prior methods.

ProStream maintains lower latency compared to full-context models.

STEM-Bench provides a new benchmark for streaming dialogue memory evaluation.

Abstract

Real-world dialogue usually unfolds as an infinite stream. It thus requires bounded-state memory mechanisms to operate within an infinite horizon. However, existing read-then-think memory is fundamentally misaligned with this setting, as it cannot support ad-hoc memory recall while streams unfold. To explore this challenge, we introduce \textbf{STEM-Bench}, the first benchmark for \textbf{ST}reaming \textbf{E}valuation of \textbf{M}emory. It comprises over 14K QA pairs in dialogue streams that assess perception fidelity, temporal reasoning, and global awareness under infinite-horizon constraints. The preliminary analysis on STEM-Bench indicates a critical textit{fidelity-efficiency dilemma}: retrieval-based methods use fragment context, while full-context models incur unbounded latency. To resolve this, we propose \textbf{ProStream}, a proactive memory framework for streaming dialogues built on a hierarchical structure. It enables ad-hoc memory recall on demand by reasoning over continuous streams with multi-granular distillation. Moreover, it employs Adaptive Spatiotemporal Optimization to dynamically optimize retention based on expected utility. It enables a bounded knowledge state for lower inference latency without sacrificing reasoning fidelity. Experiments show ProStream delivers higher reasoning fidelity than prior baselines while maintaining substantially lower latency than full-context alternatives.