Elamur Research⌕
☰
home / papers / 2026 / MAGMA: A Multi-Graph based Agentic Memory Architecture for AI Agents

MAGMA: A Multi-Graph based Agentic Memory Architecture for AI Agents

Paper · 2026 · Dongming Jiang, Yi Li, Guanpeng Li et al.

Summary

Memory-Augmented Generation (MAG) lets LLM agents keep information beyond their context window. Most current memory systems, however, put everything into one store and fetch items by embedding similarity. As a result, time order, cause-and-effect links and entity identity get mixed together, and retrieval often misses what a question actually needs.

MAGMA addresses this by recording each event node in four separate relation graphs (semantic, temporal, causal and entity), alongside a vector database. At query time, the system works in three steps:

  • an intent-aware router picks which graphs matter;
  • an adaptive traversal policy walks those graphs from anchor nodes and prunes irrelevant regions;
  • the resulting subgraphs are merged into a compact context.

Writing is split into a fast ingestion path and a slower background consolidation step, where an LLM infers causal edges.

On LoCoMo and LongMemEval, MAGMA outperforms earlier agentic memory systems, answers faster and keeps token use moderate. The work matters because retrieval paths become explicit and inspectable, and the experiments show that each relation type adds something to long-horizon reasoning.

Key points

  • The architecture has three layers: a query process (Intent-Aware Router, Adaptive Topological Retrieval, Context Synthesizer), a data layer (a time-variant directed multigraph plus a vector index), and a dual-stream write path (Synaptic Ingestion and Asynchronous Consolidation).
  • MAGMA has the lowest reported query latency, 1.47 s, about 40% faster than A-MEM, the next-best retrieval baseline, at roughly 3.37k tokens per query.
  • In the leave-one-out ablation, the full system reaches an LLM-judge score of 0.700, the setting the single-graph ablation reports for LoCoMo.
    • Removing the adaptive policy causes the largest drop, to 0.637.
    • Removing causal links gives 0.644, and removing the temporal backbone gives 0.647.
    • Removing entity links gives 0.666.
  • In the single-graph tests on LoCoMo, no variant exceeds 0.60 overall.
    • Causal Only scores 0.590 overall.
    • Entity Only scores 0.531 overall.
    • Temporal Only reaches 0.620 on temporal questions.
  • The authors note several limitations: graph quality depends on LLM extraction accuracy, the system carries extra storage and engineering overhead, and evaluation covers only conversational benchmarks. Code is released at github.com/FredJiang0324/MAGMA.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓The University of Texas at DallasRichardson · United States61 works · 5,614 citations🎓University of FloridaGainesville · United States33 works · 364 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksComputer Vision and Pattern Recognition · 2,869 worksInformation Systems · 1,437 works

Cite

Cite as Dongming Jiang, Yi Li, Guanpeng Li and Bingzhe Li. 2026. MAGMA: A Multi-Graph based Agentic Memory Architecture for AI Agents. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.1709

@inproceedings{jiang2026magma,
  title = {{MAGMA: A Multi-Graph based Agentic Memory Architecture for AI Agents}},
  author = {Dongming Jiang, Yi Li, Guanpeng Li and Bingzhe Li},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.1709},
  url = {https://doi.org/10.18653/v1/2026.acl-long.1709}
}

References

48 references, 19 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, and 1 others. 2023. Gpt-4 technical report. arXiv preprint arXiv:2303.08774.
  2. Iz Beltagy, Matthew E Peters, and Arman Cohan. 2020a. Longformer: The long-document transformer. arXiv preprint arXiv:2004.05150.
  3. Iz Beltagy, Matthew E Peters, and Arman Cohan. 2020b. Longformer: The long-document transformer. arXiv preprint arXiv:2004.05150. for retrieval-augmented generation serving. In Proceedings of the 52nd Annual International Symposium on Computer Architecture, pages 974–989.
  4. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, and 1 others. 2020. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901.
  5. Ziyan Jiang, Xueguang Ma, and Wenhu Chen. 2024. Longrag: Enhancing retrieval-augmented generation with long-context llms. arXiv preprint arXiv:2406.15319. — [arXiv:2406.15319](https://arxiv.org/abs/2406.15319) find
  6. Prateek Chhikara, Dev Khant, Saket Aryan, Taranjeet Singh, and Deshraj Yadav. 2025. Mem0: Building production-ready ai agents with scalable long-term memory. arXiv preprint arXiv:2504.19413.
  7. Gordon V. Cormack, Charles L A Clarke, and Stefan Buettcher. 2009. Reciprocal rank fusion outperforms condorcet and individual rank learning methods. In Proceedings of the 32nd International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR ’09, page 758–759, New York,…
  8. Darren Edge, Ha Trinh, Newman Cheng, Joshua Bradley, Alex Chao, Apurva Mody, Steven Truitt, Dasha Metropolitansky, Robert Osazuwa Ness, and Jonathan Larson. 2024a. From local to global: A graph rag approach to query-focused summarization. arXiv preprint arXiv:2404.16130. find
  9. Darren Edge, Ha Trinh, Newman Cheng, Joshua Bradley, Alex Chao, Apurva Mody, Steven Truitt, Dasha Metropolitansky, Robert Osazuwa Ness, and Jonathan Larson. 2024b. From local to global: A graph rag approach to query-focused summarization. arXiv preprint arXiv:2404.16130. find
  10. Zhijing Jin, Yuen Chen, Felix Leeb, Luigi Gresele, Ojasv Kamal, Zhiheng Lyu, Kevin Blin, Fernando Gonzalez Adauto, Max Kleiman-Weiner, Mrinmaya Sachan, and 1 others. 2023. Cladder: Assessing causal reasoning in language models. Advances in Neural Information Processing Systems, 36:31038– 31065. find
  11. Jeff Johnson, Matthijs Douze, and Hervé Jégou. 2019. Billion-scale similarity search with gpus. IEEE Transactions on Big Data, 7(3):535–547.
  12. Jiazheng Kang, Mingming Ji, Zhe Zhao, and Ting Bai. 2025a. Memory os of ai agent. arXiv preprint arXiv:2506.06326.
  13. Jiazheng Kang, Mingming Ji, Zhe Zhao, and Ting Bai. 2025b. Memory os of ai agent. arXiv preprint arXiv:2506.06326.
  14. Jikun Kang, Wenqi Wu, Filippos Christianos, Alex J Chan, Fraser Greenlee, George Thomas, Marvin Pur-torab, and Andy Toulis. 2025c. Lm2: Large memory models. arXiv preprint arXiv:2502.06049. — [arXiv:2502.06049](https://arxiv.org/abs/2502.06049) find
  15. Urvashi Khandelwal, He He, Peng Qi, and Dan Jurafsky. 2018. Sharp nearby, fuzzy far away: How neural language models use context. arXiv preprint arXiv:1805.04623. find
  16. Bernal Jiménez Gutiérrez, Yiheng Shu, Weijian Qi, Sizhe Zhou, and Yu Su. 2025. From rag to memory: Non-parametric continual learning for large language models. arXiv preprint arXiv:2502.14802. find
  17. Emre Kiciman, Robert Ness, Amit Sharma, and Chenhao Tan. 2023. Causal reasoning and large language models: Opening a new frontier for causality. Transactions on Machine Learning Research. find
  18. Yuyang Hu, Shichun Liu, Yanwei Yue, Guibin Zhang, Boyang Liu, Fangyi Zhu, Jiahang Lin, Honglin Guo, Shihan Dou, Zhiheng Xi, and 1 others. 2025. Memory in the age of ai agents. arXiv preprint arXiv:2512.13564. — [arXiv:2512.13564](https://arxiv.org/abs/2512.13564) find
  19. Dharshan Kumaran, Demis Hassabis, and James L Mc-Clelland. 2016. What learning systems do intelligent agents need? complementary learning systems theory updated. Trends in cognitive sciences, 20(7):512– 534. find
  20. Le Huang, Hengzhi Lan, Zijun Sun, Chuan Shi, and Ting Bai. 2024. Emotional rag: Enhancing role-playing agents through emotional retrieval. In 2024 IEEE International Conference on Knowledge Graph (ICKG), pages 120–127. IEEE. find
  21. Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, and 1 others. 2020. Retrieval-augmented generation for knowledge-intensive nlp tasks. Advances in neural information processing systems, 33:9459–…
  22. Dongming Jiang, Yi Li, Songtao Wei, Jinxin Yang, Ayushi Kishore, Alysa Zhao, Dingyi Kang, Xu Hu, Feng Chen, Qiannan Li, and 1 others. 2026. Anatomy of agentic memory: Taxonomy and empirical analysis of evaluation and system limitations. arXiv preprint arXiv:2602.19320. — [arXiv:2602.19320](https://arxiv.org/abs/2602.19320) find
  23. Wenqi Jiang, Suvinay Subramanian, Cat Graves, Gustavo Alonso, Amir Yazdanbakhsh, and Vidushi Dadu. 2025. Rago: Systematic performance optimization Yi Li, Lianjie Cao, Faraz Ahmed, Puneet Sharma, and Bingzhe Li. 2026. Hippocampus: An efficient and scalable memory module for agentic ai. arXiv preprint… — [arXiv:2602.13594](https://arxiv.org/abs/2602.13594) find
  24. Zhiyu Li, Shichao Song, Chenyang Xi, Hanyu Wang, Chen Tang, and 1 others. 2025. Memos: A memory os for ai system. arXiv preprint arXiv:2507.03724. — [arXiv:2507.03724](https://arxiv.org/abs/2507.03724) find
  25. Shuhang Lin, Zhencan Peng, Lingyao Li, Xiao Lin, Xi Zhu, and Yongfeng Zhang. 2025. Cache mechanism for agent rag systems. arXiv preprint arXiv:2511.02919. — [arXiv:2511.02919](https://arxiv.org/abs/2511.02919) find
  26. Lei Liu, Xiaoyan Yang, Yue Shen, Binbin Hu, Zhiqiang Zhang, Jinjie Gu, and Guannan Zhang. 2023. Thinkin-memory: Recalling and post-thinking enable llms with long-term memory. arXiv preprint arXiv:2311.08719. — [arXiv:2311.08719](https://arxiv.org/abs/2311.08719) find
  27. Nelson F Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang. 2024. Lost in the middle: How language models use long contexts. Transactions of the Association for Computational Linguistics, 12:157–173.
  28. Adyasha Maharana, Dong-Ho Lee, Sergey Tulyakov, Mohit Bansal, Francesco Barbieri, and Yuwei Fang. 2024. Evaluating very long-term conversational memory of llm agents. arXiv preprint arXiv:2402.17753. find
  29. Jiayan Nan, Wenquan Ma, Wenlong Wu, and Yize Chen. 2025. Nemori: Self-organizing agent memory inspired by cognitive science. arXiv preprint arXiv:2508.03341. — [arXiv:2508.03341](https://arxiv.org/abs/2508.03341) find
  30. Charles Packer, Vivian Fang, Shishir_G Patil, Kevin Lin, Sarah Wooders, and Joseph_E Gonzalez. 2023. Memgpt: Towards llms as operating systems. find
  31. Shirui Pan, Linhao Luo, Yufei Wang, Chen Chen, Jiapu Wang, and Xindong Wu. 2024. Unifying large language models and knowledge graphs: A roadmap. IEEE Transactions on Knowledge and Data Engineering, 36(7):3580–3599.
  32. Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics, pages 311–318, Philadelphia, Pennsylvania, USA. Association for Computational…
  33. Joon Sung Park, Joseph O’Brien, Carrie Jun Cai, Meredith Ringel Morris, Percy Liang, and Michael S Bernstein. 2023. Generative agents: Interactive simulacra of human behavior. In Proceedings of the 36th annual acm symposium on user interface software and technology, pages 1–22. find
  34. Ofir Press, Noah A Smith, and Mike Lewis. 2021a. Train short, test long: Attention with linear biases enables input length extrapolation. arXiv preprint arXiv:2108.12409. global memory-enhanced retrieval augmentation. In Proceedings of the ACM on Web Conference 2025, pages 2366–2377. find
  35. Preston Rasmussen, Pavlo Paliychuk, Travis Beauvais, Jack Ryan, and Daniel Chalef. 2025. Zep: a temporal knowledge graph architecture for agent memory. arXiv preprint arXiv:2501.13956. find
  36. Somin Wadhwa, Silvio Amir, and Byron C Wallace. 2023. Revisiting relation extraction in the era of large language models. In Proceedings of the conference. association for computational linguistics. meeting, volume 2023, page 15566. find
  37. Yu Wang and Xi Chen. 2025. Mirix: Multi-agent memory system for llm-based agents. arXiv preprint arXiv:2507.07957. — [arXiv:2507.07957](https://arxiv.org/abs/2507.07957) find
  38. Zheng Wang, Shu Teo, Jieer Ouyang, Yongjun Xu, and Wei Shi. 2024. M-rag: Reinforcing large language model performance through retrieval-augmented generation with multiple partitions. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), p… find
  39. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, and 1 others. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824– 24837.
  40. Di Wu, Hongwei Wang, Wenhao Yu, Yuwei Zhang, Kai-Wei Chang, and Dong Yu. 2024. Longmemeval: Benchmarking chat assistants on long-term interactive memory. arXiv preprint arXiv:2410.10813.
  41. Zhiheng Xi, Wenxiang Chen, Xin Guo, Wei He, Yiwen Ding, Boyang Hong, Ming Zhang, Junzhe Wang, Senjie Jin, Enyu Zhou, and 1 others. 2025. The rise and potential of large language model based agents: A survey. Science China Information Sciences, 68(2):121101.
  42. Shijia Xu, Zhou Wu, Xiaolong Jia, Yu Wang, Kai Liu, and April Xiaowen Dong. 2026. Self-correcting rag: Enhancing faithfulness via mmkp context selection and nli-guided mcts. Preprint, arXiv:2604.10734.
  43. Wujiang Xu, Zujie Liang, Kai Mei, Hang Gao, Juntao Tan, and Yongfeng Zhang. 2025. A-mem: Agentic memory for llm agents. arXiv preprint arXiv:2502.12110.
  44. Ofir Press, Noah A Smith, and Mike Lewis. 2021b. Train short, test long: Attention with linear biases enables input length extrapolation. arXiv preprint arXiv:2108.12409. find
  45. Yiming Zeng, Wanhao Yu, Zexin Li, Tao Ren, Yu Ma, Jinghan Cao, Xiyan Chen, and Tingting Yu. 2025. Bridging the editing gap in llms: Fineedit for precise and targeted text modifications. Findings of the Association for Computational Linguistics: EMNLP 2025, pages 2193–2206. find
  46. Hongjin Qian, Zheng Liu, Peitian Zhang, Kelong Mao, Defu Lian, Zhicheng Dou, and Tiejun Huang. 2025. Memorag: Boosting long context processing with Zhuosheng Zhang, Yao Yao, Aston Zhang, Xiangru Tang, Xinbei Ma, Zhiwei He, Yiming Wang, Mark Gerstein, Rui Wang, Gongshen Liu, and 1 others. 2025. Ignit… find
  47. Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric Xing, and 1 others. 2023. Judging llm-as-a-judge with mt-bench and chatbot arena. Advances in neural information processing systems, 36:46595–46623.
  48. Wanjun Zhong, Lianghong Guo, Qiqi Gao, He Ye, and Yanlin Wang. 2024. Memorybank: Enhancing large language models with long-term memory. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 19724–19731. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/jiang-2026-magma-multi-graph-based-agentic/