Language agents manage memory by separating stored facts from used evidence
What Should an Agent Forget? Separating What Is Stored from What Is Used
Artificial Intelligence
Summary
Sometimes, an AI assistant needs to remember facts over time, but not all facts should be used in every answer. For example, old facts might be wrong now but still important for history questions. The researchers created a system called RD-Forget that keeps all facts stored but only uses the relevant and current ones when answering. This system groups related facts and knows when to ignore outdated information depending on the question. This helps AI give more accurate and flexible answers.
What this means in practice
- •For chatbot developers: Build conversational agents that recall past interactions while ignoring outdated facts to improve answer accuracy.
- •For customer support teams: Maintain up-to-date product information in AI helpers while keeping access to historical data for reference in long-running cases.
Authors
Yuhang Li, Yuchen Li
Abstract
Persistent language agents need stored experience to remain available across time, while each answer requires evidence suited to a particular question. A superseded fact can mislead a current-state answer and still be essential for a historical query. We present RD-Forget, a training-free framework that separates what an agent stores from what it uses. A retained source archive preserves observations, and a query-conditioned memory view controls their influence on the current answer. A frozen language-model curator extracts relevant evidence, groups facts into semantic slots, and preserves the relations needed for multi-hop reasoning. Same-slot replacement links suppress superseded values in current-state contexts, while intent-aware retrieval makes earlier evidence eligible again. A rate-distortion formulation guides construction of the answer-time view within a memory budget. Experiments span conversational memory, knowledge updating, fact consolidation, long-context reasoning, and personalization under a shared answering pipeline. The results associate accurate answers with both query-relevant evidence construction and control over obsolete alternatives. Configurations without forgetting or query conditioning have the largest score deficits, while slot grouping, historical access, and relation preservation contribute complementary functions. Retaining history while selectively controlling its use offers a practical way to accommodate changing facts and future questions.