Don't Scroll Back: Missing-Evidence Memory for Streaming Dialogue Summarization
2026-08-10 • Computation and Language
Computation and LanguageArtificial Intelligence
AI summaryⓘ
The authors address the problem of summarizing ongoing conversations when only part of the full discussion is visible at a time. They focus on a system that uses a limited memory to recall important past information needed to understand the current part. Their method, called ReMEMBER, selectively retrieves and stores key details that fill in missing context so summaries make sense. Tests on very long dialogues show that ReMEMBER better captures relevant history and improves how complete the summaries are compared to other methods with the same memory limits.
streaming dialogue summarizationdialogue historymemory retrievalcontext dependenciessummary generationevidence recallevaluation protocolmemory budgetgap resolutionlong-range dialogue
Authors
Hyangsuk Min, Hwanjun Song
Abstract
Users of modern platforms repeatedly need summaries of recent dialogue, but the window rarely contains enough context to be interpreted on its own. We formalize this setting as streaming dialogue summarization, where a system must summarize a current window using selective memory from an unbounded history under a fixed budget. We show that the central challenge is not how much history is accessed, but whether memory recovers the evidence that the current window presupposes. We construct a benchmark and evaluation protocol that separately assesses whether memory contains gap-resolving evidence and whether the generated summary reflects it. We propose ReMEMBER, a missing-evidence memory framework that conditions retrieval on unresolved window dependencies and refines retrieved chunks into evidence-dense memory under a fixed budget. Experiments on dialogues with histories up to 160K tokens show that ReMEMBER improves memory recall and gap-resolution completeness over memory construction baselines under the same budget.