Grounding Memory Summarization in Utility Intent
MemSuit improves memory summarization by self-distilling query-conditioned utility awareness into raw-conversation entries and decomposing blocks to prevent collateral erasure, boosting answer quality across query types.
Published Aug 20, 2026
Only vote on papers you've read. Sign in with GitHub to vote.
MemSuit's utility-conditioned self-distillation and decomposed blocks decisively beat faithfulness-based memory, yet its gains remain tied to an unvalidated synthetic failure mode, missing adversarial query splits, block-count ablations, and proof that the contrastive retriever outperforms raw-conversation embeddings…
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.