aiqbench analysis
This paper proposes modeling long-term memory in personalized AI as a dynamical state space with locally heterogeneous geometry, moving beyond static latent spaces to capture temporal trajectories and user drift. By treating memory access as trajectory-conditioned reconstruction rather than simple nearest-neighbor lookup, the approach better aligns computational abstractions with the continuous nature of event streams. This shift represents a significant methodological advancement in how AI systems encode and retrieve personalization data over extended horizons.
Memory Has Geometry: Non-Uniform Geometric Memory for Long-Horizon Personalized AI
arXiv:2609.17969v1 Announce Type: new Abstract: Long-term memory is becoming a core substrate for personalized AI, yet most systems still represent personalization as discrete records in a largely static latent space, accessed under one global similarity notion. For data mining, this creates a mismatch: the evidence is a temporal event stream, while the dominant abstraction is a searchable record set. We argue that long-horizon personalization should instead model memory as a user-specific dynamical state space with locally heterogeneous geometry. Geometry here is a computational language, not a literal claim about cognition: it captures stable versus volatile regions, variable-rate drift, heterogeneous neighborhoods, and uncertainty about current user state. Profiles and isolated events remain useful as points, but interaction, feedback, and elapsed time induce trajectories. Memory access then becomes trajectory-conditioned reconstruction of the relevant user state, not only nearest-neighbor lookup.
Abstract:Long-term memory is becoming a core substrate for personalized AI, yet most systems still represent personalization as discrete records in a largely static latent space, accessed under one global similarity notion. For data mining, this creates a mismatch: the evidence is a temporal event stream, while the dominant abstraction is a searchable record set. We argue that long-horizon personalization should instead model memory as a user-specific dynamical state space with locally heterogeneous geometry. Geometry here is a computational language, not a literal claim about cognition: it captures stable versus volatile regions, variable-rate drift, heterogeneous neighborhoods, and uncertainty about current user state. Profiles and isolated events remain useful as points, but interaction, feedback, and elapsed time induce trajectories. Memory access then becomes trajectory-conditioned reconstruction of the relevant user state, not only nearest-neighbor lookup.
Accepted to the IEEE ICDM 2026 (BlueSky). 6 pages, 2 figures, and 2 tables
Artificial Intelligence (cs.AI)
arXiv:2609.17969 [cs.AI]
(or arXiv:2609.17969v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2609.17969 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 240 words
Related coverage
WorldPack: Dynamic Frame Compression for Long-context Video World Modeling
arXiv · Jul 23, 2026
What's Continual Learning, and Why Might We Expect To See It In Advanced LLM Agents?
LessWrong · Jun 12, 2026
[AINews] not much happened today
Latent Space · Sep 18, 2026
AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
arXiv · Sep 7, 2026