Multi-goal agent memory improves by writing goal-specific summaries
Interpreting at Write Time: A Policy Ablation for Multi-Goal Agent Memory
Artificial Intelligence
Summary
When a digital assistant remembers past events, it can’t keep everything and must make summaries. What gets saved depends on what the assistant thinks is important for the user’s goals. The paper studies three ways to summarize memories: ignoring goals, mixing all goals into one summary, or making separate summaries for each goal. They found that making individual summaries for each goal works best for accuracy and relevance, but only if the summary is later used for that specific goal. The authors show that focusing on goals when writing summaries helps preserve useful information.
What this means in practice
- •For conversational ai developers: Create digital assistants that keep goal-specific memory summaries to improve relevance and accuracy in long conversations when multiple user goals are present.$Commercial implications: Enables more precise and personalized AI assistants for customer support or personal use by tailoring memory storage to user goals.
- •For knowledge management teams: Implement goal-focused memory summaries to optimize storage and retrieval in systems tracking multiple projects or objectives simultaneously.
Authors
Albert Sadowski, Jarosław A. Chudziak
Abstract
A long-running assistant cannot keep everything it has seen, so it summarises. Summarising is not neutral: what is kept is chosen against some notion of what the record is for, and that choice is made once, before anyone knows which of the user's standing goals will ask. Goals rarely disagree about what happened. They disagree about which parts of it were worth the space. Once the history is too long to re-read, the summary replaces the stream, and whatever it left out is gone. We ask what a memory should summarise for when it serves several standing goals at once. Three policies answer differently: summarise with no goal in view, write one summary covering every goal, or write one summary per goal and read them together. We compare them across several models and event streams, holding the read step fixed so that only the write differs. The goals do pull apart: summaries written for different goals overlap each other less than a summary overlaps a rewrite of itself. Per-goal summaries win on relevance, completeness and accuracy, and the all-goal summary loses even to the neutral one written at a fraction of its budget. Interpreting at write pays off, but only for the goal that later asks.