Memory storing changes how long user facts seem to hold true
Memory Consolidation Flattens the Temporal Shape of User Facts
Computation and Language
Summary
When computer memory systems save information about users, they often shorten sentences and lose clues about whether things are still happening. For example, ‘I am driving a Peugeot’ becomes ‘The user drives a Peugeot,’ which no longer shows that the driving is current. The authors studied how this change, called aspectual flattening, happens consistently in memory systems. They also found that this loss of timing cues affects how later systems decide to act on stored facts, sometimes leading to less cautious behavior.
What this means in practice
- •For dialog system developers: Improve memory modules to preserve temporal cues so responses better reflect whether user facts remain valid.
- •For digital assistant engineers: Adjust fact storage to reduce mistakes where assistants act without confirming if a fact still applies to the user.
Authors
Sugam Panthi, Muhaiminul Yeamin, Siyan Luo, Rabab Abdelfattah
Abstract
Long-term memory systems turn conversations into short stored notes. A note can keep a user fact while losing evidence about whether the fact still holds. For example, "I am driving a Peugeot" can become "The user drives a Peugeot," which drops the cue that the activity is ongoing. We call this aspectual flattening and measure it with LAPSE, a benchmark of matched user statements that differ only in temporal form. We find that memory writers flatten aspect selectively. Three writer models flattened the progressive statement but kept its simple-present match in 244 of 381 pairs, never the reverse. The asymmetry holds in all 11 model configurations tested and in the installed pipelines mem0, Graphiti, and Letta. The lost cue matters to later readers. In exploratory tests, changing only the stored verb shifted all three readers' estimates that a fact still holds. When readers could ask the user before acting, two of three acted without asking more often on flattened notes. Our planned memory-use task could not detect this, because readers there acted on almost every stored fact, even expired ones. Memory writing can thus remove evidence that later models use to decide whether to act.