AssoMem: Scalable Memory QA with Multi-Signal Associative Retrieval

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Kai, Zhang, Xinyuan, Ahmed, Ejaz, Jiang, Hongda, Kumar, Caleb, Sun, Kai, Lin, Zhaojiang, Sharma, Sanat, Oraby, Shereen, Colak, Aaron, Aly, Ahmed, Kumar, Anuj, Liu, Xiaozhong, Dong, Xin Luna
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914089443262464
author Zhang, Kai
Zhang, Xinyuan
Ahmed, Ejaz
Jiang, Hongda
Kumar, Caleb
Sun, Kai
Lin, Zhaojiang
Sharma, Sanat
Oraby, Shereen
Colak, Aaron
Aly, Ahmed
Kumar, Anuj
Liu, Xiaozhong
Dong, Xin Luna
author_facet Zhang, Kai
Zhang, Xinyuan
Ahmed, Ejaz
Jiang, Hongda
Kumar, Caleb
Sun, Kai
Lin, Zhaojiang
Sharma, Sanat
Oraby, Shereen
Colak, Aaron
Aly, Ahmed
Kumar, Anuj
Liu, Xiaozhong
Dong, Xin Luna
contents Accurate recall from large scale memories remains a core challenge for memory augmented AI assistants performing question answering (QA), especially in similarity dense scenarios where existing methods mainly rely on semantic distance to the query for retrieval. Inspired by how humans link information associatively, we propose AssoMem, a novel framework constructing an associative memory graph that anchors dialogue utterances to automatically extracted clues. This structure provides a rich organizational view of the conversational context and facilitates importance aware ranking. Further, AssoMem integrates multi-dimensional retrieval signals-relevance, importance, and temporal alignment using an adaptive mutual information (MI) driven fusion strategy. Extensive experiments across three benchmarks and a newly introduced dataset, MeetingQA, demonstrate that AssoMem consistently outperforms SOTA baselines, verifying its superiority in context-aware memory recall.
format Preprint
id arxiv_https___arxiv_org_abs_2510_10397
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle AssoMem: Scalable Memory QA with Multi-Signal Associative Retrieval
Zhang, Kai
Zhang, Xinyuan
Ahmed, Ejaz
Jiang, Hongda
Kumar, Caleb
Sun, Kai
Lin, Zhaojiang
Sharma, Sanat
Oraby, Shereen
Colak, Aaron
Aly, Ahmed
Kumar, Anuj
Liu, Xiaozhong
Dong, Xin Luna
Computation and Language
Accurate recall from large scale memories remains a core challenge for memory augmented AI assistants performing question answering (QA), especially in similarity dense scenarios where existing methods mainly rely on semantic distance to the query for retrieval. Inspired by how humans link information associatively, we propose AssoMem, a novel framework constructing an associative memory graph that anchors dialogue utterances to automatically extracted clues. This structure provides a rich organizational view of the conversational context and facilitates importance aware ranking. Further, AssoMem integrates multi-dimensional retrieval signals-relevance, importance, and temporal alignment using an adaptive mutual information (MI) driven fusion strategy. Extensive experiments across three benchmarks and a newly introduced dataset, MeetingQA, demonstrate that AssoMem consistently outperforms SOTA baselines, verifying its superiority in context-aware memory recall.
title AssoMem: Scalable Memory QA with Multi-Signal Associative Retrieval
topic Computation and Language
url https://arxiv.org/abs/2510.10397