Metraipolis › archive › dc_57e84902ddcf

Disentangling Long-Term Memory via Latent Neuro-Symbolic Reasoning — arXiv:2609.18461

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.18461
License (as archived)
CC0-1.0
Archived
2026-10-15T13:49:57.048Z
SHA-256 of the archived text
ee0671115120a957325d213bc373ac7af72af8fa79f8e6b66d3df5d604c900e5
Stable URL of this page
https://metraipolis.com/documento/dc_57e84902ddcf

How to cite

Disentangling Long-Term Memory via Latent Neuro-Symbolic Reasoning — arXiv:2609.18461 — arXiv (en), https://arxiv.org/abs/2609.18461 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_57e84902ddcf (Metraipolis archive dc_57e84902ddcf, sha256 ee0671115120a957325d213bc373ac7af72af8fa79f8e6b66d3df5d604c900e5)
BibTeX
@misc{dc57e84902ddcf,
  title = {Disentangling Long-Term Memory via Latent Neuro-Symbolic Reasoning — arXiv:2609.18461},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.18461},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_57e84902ddcf; sha256 ee0671115120a957325d213bc373ac7af72af8fa79f8e6b66d3df5d604c900e5},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.18461. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Personalized agents are required to reason over long-term history interactions to infer both explicit preferences and implicit behavioral evidence. While early flat retrieval methods score memory fragments independently and neglect the distributed information, current structured memory frameworks rely on query-agnostic static graphs that fail to capture the context-dependent relations. Crucially, raw textual memories are inherently entangled and noisy, making fine-grained personalization and cross-session reasoning computationally prohibitive. To this end, we present LGM, a novel neuro-symbolic framework that shifts long-term memory disentanglement into a continuous latent space. Specifically, (i) instead of persisting fixed graphs, we design a tailored latent graph construction with a sparse autoencoder. Subject to each query, it maps historical interactions into latent memory nodes and disentangles the memory traces into sparse concept activations, dynamically synthesizing query-aware relational edge weights. (ii) A graph encoder then treats the query embedding as a conditioning preference to direct non-linear message passing across the task-specific latent subgraph. This yields a highly expressive memory representation for effective activations. Extensive experiments on long-term personalization benchmarks demonstrate that LGM significantly outperforms state-of-the-art baselines in capturing both explicit and implicit preferences while enabling personalized responses.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-17 in cs.CL).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Computation and Language (cs.CL); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.18461 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-15T13:48:24.420Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-17?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).