Elamur Research⌕
☰
home / papers / 2026 / Grounding Agent Memory in Contextual Intent

Grounding Agent Memory in Contextual Intent

Paper · 2026 · Ruozhen Yang, Yucheng Jiang, Yueqi Jiang et al.

Summary

Agents working on long, goal-driven tasks often meet the same entities and facts many times under different goals. A hotel price, for example, can depend on which day is being planned. Standard memory systems then pull up evidence that looks relevant but belongs to the wrong context. The authors present STITCH (Structured Intent Tracking in Contextual History), a memory system based on Event Structure Theory. It labels every trajectory step online with a "contextual intent" made of three parts: a thematic scope for the current goal segment, an event type for the action performed, and the key entity types that matter. No fixed ontology is needed. At query time, stored snippets are filtered and ranked by how well their intent matches the current step, so similar but mismatched history is pushed aside. The authors also release CAME-Bench, a synthetic benchmark with interleaved interactions that do not follow strict turn-taking. On CAME-Bench and LongMemEval, STITCH beats the strongest baseline by 35.6%, and the advantage grows on longer trajectories. This makes intent-based indexing a promising basis for dependable long-horizon agent memory.

Key points

  • Contextual intent tuple: each step gets a thematic scope, an event type and a set of key entity types.
    • The scope is induced by an LLM over a sliding window and stays fixed until the goal changes.
    • Event labels come from a vocabulary that grows over time and is periodically merged.
  • CAME-Bench (Context-Aware Agent Memory Evaluation Benchmark):
    • covers multiple domains, including Travel Planning and Debate;
    • is split by length into Small, Medium and Large subsets;
    • asks four question types: incremental memory revision, context-aware factual recall, multi-hop reasoning and information synthesis.
  • Main result: STITCH reaches state-of-the-art results on CAME-Bench and LongMemEval, with a 35.6% gain over the best baseline.
  • Backbone comparison (gpt-4o-mini, gpt-4.1-mini):
    • STITCH outperforms SeCom on every partition.
    • With gpt-4.1-mini, a full-context baseline is stronger on Small, but STITCH overtakes it on Medium and Large.
  • Limitations: building the intent tuples requires several LLM calls per step, so ingestion costs more than embedding-only methods. The label space is updated in buffers of N=50 steps.

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

🎓Stanford UniversityStanford · United States434 works · 314,419 citations🎓University of Illinois Urbana-ChampaignUrbana · United States184 works · 69,372 citations

 Fields, topics and concepts

📚 Shelves 5The 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 worksInformation Systems · 1,437 works

Cite

Cite as Ruozhen Yang, Yucheng Jiang, Yueqi Jiang, Priyanka Kargupta, Yunyi Zhang and Jiawei Han. 2026. Grounding Agent Memory in Contextual Intent. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-acl.584

@inproceedings{yang2026grounding,
  title = {{Grounding Agent Memory in Contextual Intent}},
  author = {Ruozhen Yang, Yucheng Jiang, Yueqi Jiang, Priyanka Kargupta, Yunyi Zhang and Jiawei Han},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.584},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.584}
}

References

26 references, 12 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Judith Avrahami and Yaakov Kareev. 1994. The emergence of events. Cognition, 53(3):239–261. find
  2. Ge Bai, Jie Liu, Xingyuan Bu, Yancheng He, Jiaheng Liu, Zhanhui Zhou, Zhuoran Lin, Wenbo Su, Tiezheng Ge, Bo Zheng, and Wanli Ouyang. 2024a. MT-bench-101: A fine-grained benchmark for evaluating large language models in multi-turn dialogues. In Proceedings of the 62nd Annual Meeting of the Associati…
  3. Yushi Bai, Shangqing Tu, Jiajie Zhang, Hao Peng, Xiaozhi Wang, Xin Lv, Shulin Cao, Jiazheng Xu, Lei Hou, Yuxiao Dong, Jie Tang, and Juanzi Li. 2024b. Longbench v2: Towards deeper understanding and reasoning on realistic long-context multitasks. arXiv preprint arXiv:2412.15204.
  4. Roger G Barker and Herbert F Wright. 1954. Midwest and the USA. Row, Peterson and Company. find
  5. Gheorghe Comanici, Eric Bieber, Mike Schaekermann, Ice Pasupat, Noveen Sachdeva, Inderjit Dhillon, Marcel Blistein, Ori Ram, Dan Zhang, Evan Rosen, and 1 others. 2025. Gemini 2.5: Pushing the frontier with advanced reasoning, multimodality, long context, and next generation agentic capabilities. arX…
  6. Fergus IM Craik and Robert S Lockhart. 1972. Levels of processing: A framework for memory research. Journal of verbal learning and verbal behavior, 11(6):671–684. find
  7. Darren Edge, Ha Trinh, Newman Cheng, Joshua Bradley, Alex Chao, Apurva Mody, Steven Truitt, Dasha Metropolitansky, Robert Osazuwa Ness, and Jonathan Larson. 2024. From local to global: A graph rag approach to query-focused summarization. find
  8. Mihail Eric, Rahul Goel, Shachi Paul, Abhishek Sethi, Sanchit Agarwal, Shuyang Gao, Adarsh Kumar, Anuj Goyal, Peter Ku, and Dilek Hakkani-Tur. 2020. MultiWOZ 2.1: A consolidated multi-domain dialogue dataset with state corrections and state tracking baselines. In Proceedings of the Twelfth Language… find
  9. 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. Preprint, arXiv:2502.14802. find
  10. Yucheng Jiang, Yijia Shao, Dekun Ma, Sina Semnani, and Monica Lam. 2024. Into the unknown unknowns: Engaged human learning through participation in language model agent conversations. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 9917–9955, Miami, F… find
  11. Vladimir Karpukhin, Barlas Oguz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih. 2020. Dense passage retrieval for open-domain question answering. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 6769–6781, Onli…
  12. 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.
  13. Adyasha Maharana, Dong-Ho Lee, Sergey Tulyakov, Mohit Bansal, Francesco Barbieri, and Yuwei Fang. 2024. Evaluating very long-term conversational memory of llm agents. Preprint, arXiv:2402.17753. find
  14. Darren Newtson, Gretchen A Engquist, and Joyce Bois. 1977. The objective basis of behavior units. Journal of Personality and social psychology, 35(12):847. find
  15. OpenAI. 2025. Introducing GPT-4.1 in the API. Accessed: 2025-11-27. find
  16. Zhuoshi Pan, Qianhui Wu, Huiqiang Jiang, Xufang Luo, Hao Cheng, Dongsheng Li, Yuqing Yang, Chin-Yew Lin, H. Vicky Zhao, Lili Qiu, and Jianfeng Gao. 2025. Secom: On memory construction and retrieval for personalized conversational agents. In The Thirteenth International Conference on Learning Represe… find
  17. Abhinav Rastogi, Xiaoxue Zang, Srinivas Sunkara, Raghav Gupta, and Pranav Khaitan. 2020. Towards scalable multi-domain conversational agents: The schema-guided dialogue dataset. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 34, pages 8689–8696.
  18. Oscar Sainz, Jon Campos, Iker García-Ferrero, Julen Etxaniz, Oier Lopez de Lacalle, and Eneko Agirre. 2023. NLP evaluation in trouble: On the need to measure LLM data contamination for each benchmark. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 10776–10787, Singap…
  19. Parth Sarthi, Salman Abdullah, Aditi Tuli, Shubh Khanna, Anna Goldie, and Christopher D. Manning. 2024. Raptor: Recursive abstractive processing for tree-organized retrieval. In International Conference on Learning Representations (ICLR).
  20. Yijia Shao, Yucheng Jiang, Theodore Kanell, Peter Xu, Omar Khattab, and Monica Lam. 2024. Assisting in writing Wikipedia-like articles from scratch with large language models. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human…
  21. Leonard Talmy. 1975. Semantics and syntax of motion. In Syntax and Semantics volume 4, pages 181–238. Brill. find
  22. Endel Tulving and Donald M Thomson. 1973. Encoding specificity and retrieval processes in episodic memory. Psychological review, 80(5):352. find
  23. Di Wu, Hongwei Wang, Wenhao Yu, Yuwei Zhang, Kai-Wei Chang, and Dong Yu. 2024. Longmemeval: Benchmarking chat assistants on long-term interactive memory. Preprint, arXiv:2410.10813.
  24. Tianbao Xie, Danyang Zhang, Jixuan Chen, Xiaochuan Li, Siheng Zhao, Ruisheng Cao, Toh Jing Hua, Zhoujun Cheng, Dongchan Shin, Fangyu Lei, Yitao Liu, Yiheng Xu, Shuyan Zhou, Silvio Savarese, Caiming Xiong, Victor Zhong, and Tao Yu. 2024. Osworld: Benchmarking multimodal agents for open-ended tasks in…
  25. 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.
  26. Jeffrey M Zacks and Barbara Tversky. 2001. Event structure in perception and conception. Psychological bulletin, 127(1):3. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/yang-2026-grounding-agent-memory-contextual-intent/