Memory summarization improved by focusing on future query needs
Grounding Memory Summarization in Utility Intent
Computation and Language
Summary
When computers remember past conversations, they usually save summaries meant to be truthful to humans, but this doesn’t always help answer future questions well. The authors show that making summaries based on the questions and answers the system expects to see helps keep the right information. They create a method called MemSuit that learns to make these useful summaries from raw conversations without needing future questions in advance. This approach also breaks information into smaller parts so important details for different questions aren’t lost. Their technique works better than others for various conversation types.
What this means in practice
- •For conversational ai developers: Create memory systems that better retain details needed to accurately answer future user queries during conversations.
- •For customer support teams: Improve automated chatbots’ ability to recall relevant past interactions that help resolve customer issues more effectively.
Authors
Zhenyu Lei, Mingjia Shi, Xingbo Fu, Haoyu He, Qi R. Wang, Jundong Li
Abstract
Existing summarizers for memory systems are typically optimized for human-facing criteria such as faithfulness, which misaligns with their true objective: preserving the evidence needed to support future queries. We show that conditioning summarization on query-answer pairs substantially improves answer quality, and that this utility-aware behavior is transferable across queries. Motivated by these findings, we propose MemSuit, a self-distillation framework in which a teacher summarizer, conditioned on observed query-answer pairs, produces utility-aware memory entries that a student learns to reproduce from the raw conversation alone. To prevent collateral erasure where conditioning on a single query-answer pair discards evidence relevant to other plausible queries, the teacher decomposes each block into multiple self-contained entries that preserve distinct query-relevant facets as independently retrievable units. To align the retriever with the compact, fact-dense style of teacher entries, we further fine-tune the embedding model with a contrastive objective supervised by teacher entries. Across a diverse suite of conversational query types, MemSuit consistently outperforms state-of-the-art baselines, confirming the value of grounding memory in downstream utility.