Elamur Research⌕
☰
home / papers / 2026 / HiGraAgent: Dual-Agent Adaptive Reasoning over Hierarchical Knowledge Graph for Open Domain Multi-hop Question Answering

HiGraAgent: Dual-Agent Adaptive Reasoning over Hierarchical Knowledge Graph for Open Domain Multi-hop Question Answering

Paper · 2026 · Hung Luu, Long S. T. Nguyen, Trung Pham et al.

Summary

Open-domain multi-hop QA is difficult for two reasons. Questions can have complex internal structure, and the supporting evidence is often spread over many passages. The authors argue that "multi-hopness" is not fixed by the question alone. It depends on how fragmented the available context is, and sequential iterative pipelines handle branching or parallel reasoning poorly. Their framework, HiGraAgent, has three parts:

  • HiGra: a two-level knowledge graph with an Entity Layer and a Passage Layer. It is built through coreference resolution, LLM-based extraction with a schema that records aliases and types, and entity merging by exact matching and LLM-assisted semantic matching.
  • HiGraRetriever: a retriever that combines semantic similarity with graph structure via Personalized PageRank. It returns both triplets and passages.
  • Seeker–Librarian protocol: two agents that decompose questions and explore several reasoning paths adaptively.

On HotpotQA, 2WikiMultihopQA and MuSiQue, the full system reaches 85.3% average accuracy, 11.7% above the best prior system. Its graphs are also more compact than those of the baselines. The modular design lets users trade cost against accuracy for each dataset.

Key points

  • The entity alignment step shrinks HiGra by 34.5%. The resulting graphs are 62.9% (vs. NodeRAG) to 83.7% (vs. HiRAG) the size of baseline graphs.
  • HiGraRetriever alone beats the strongest graph-based single-step baseline by an average of 3.7%. It uses only about 4,107 tokens per query.
  • The Seeker–Librarian Dual-Agent scheme reaches 83.0% average accuracy even with simple retrieval, which is 9.4% above the strongest baseline.
  • The full HiGraAgent needs about 26.1 s and 25,119 tokens per query. This is still fewer tokens than HiRAG (26,399) or DualRAG (32,342).
  • On the entity-centric 2WikiMultihopQA, graph retrieval adds 5.3 absolute points over the non-graph Dual-Agent. Code and data are released at github.com/headinthecloud6453/higra_agent.

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

🎓Vietnam National University Ho Chi Minh CityHo Chi Minh City · Vietnam21 works · 14 citations

Also on the byline: Ho Chi Minh City University of Technology.

 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

🧭 Topics 3Research subjects and languages the work is about.Reasoning · 1,426 worksQuestion answering · 441 worksKnowledge graphs & ontologies · 192 works

Cite

Cite as Hung Luu, Long S. T. Nguyen, Trung Pham, Hieu Pham and Tho Quan. 2026. HiGraAgent: Dual-Agent Adaptive Reasoning over Hierarchical Knowledge Graph for Open Domain Multi-hop Question Answering. Findings of the Association for Computational Linguistics: EACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-eacl.62

@inproceedings{luu2026higraagent,
  title = {{HiGraAgent: Dual-Agent Adaptive Reasoning over Hierarchical Knowledge Graph for Open Domain Multi-hop Question Answering}},
  author = {Hung Luu, Long S. T. Nguyen, Trung Pham, Hieu Pham and Tho Quan},
  booktitle = {Findings of the Association for Computational Linguistics: EACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-eacl.62},
  url = {https://doi.org/10.18653/v1/2026.findings-eacl.62}
}

References

23 references, 8 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Rong Cheng, Jinyi Liu, Yan Zheng, Fei Ni, Jiazhen Du, Hangyu Mao, Fuzheng Zhang, Bo Wang, and Jianye Hao. 2025. Dualrag: A dual-process approach to integrate reasoning and retrieval for multi-hop question answering. Preprint, arXiv:2504.18243.
  2. Darren Edge, Ha Trinh, Newman Cheng, Joshua Bradley, Alex Chao, Apurva Mody, Steven Truitt, Dasha Metropolitansky, Robert Osazuwa Ness, and Jonathan Larson. 2025. From Local to Global: A Graph RAG Approach to Query-Focused Summarization. Preprint, arXiv:2404.16130. find
  3. Shengbo Gong, Xianfeng Tang, Carl Yang, and Wei jin. 2025. Beyond chunks and graphs: Retrieval-augmented generation through triplet-driven thinking. Preprint, arXiv:2508.02435. find
  4. Jiawei Gu, Xuhui Jiang, Zhichao Shi, Hexiang Tan, Xuehao Zhai, Chengjin Xu, Wei Li, Yinghan Shen, Shengjie Ma, Honghao Liu, Saizhuo Wang, Kun Zhang, Yuanzhuo Wang, Wen Gao, Lionel Ni, and Jian Guo. 2025. A survey on llm-as-a-judge. Preprint, arXiv:2411.15594. find
  5. Bernal Jiménez Gutiérrez, Yiheng Shu, Yu Gu, Michihiro Yasunaga, and Yu Su. 2025a. HippoRAG: Neurobiologically Inspired Long-Term Memory for Large Language Models. Preprint, arXiv:2405.14831.
  6. Bernal Jiménez Gutiérrez, Yiheng Shu, Weijian Qi, Sizhe Zhou, and Yu Su. 2025b. From RAG to Memory: Non-Parametric Continual Learning for Large Language Models. Preprint, arXiv:2502.14802. find
  7. Xanh Ho, Anh-Khoa Duong Nguyen, Saku Sugawara, and Akiko Aizawa. 2020. Constructing A Multi-hop QA Dataset for Comprehensive Evaluation of Reasoning Steps. In Proceedings of the 28th International Conference on Computational Linguistics, pages 6609–6625, Barcelona, Spain (Online). International Comm…
  8. Haoyu Huang, Yongfeng Huang, Junjie Yang, Zhenyu Pan, Yongqiang Chen, Kaili Ma, Hongzhi Chen, and James Cheng. 2025. Hirag: Retrieval-augmented generation with hierarchical knowledge. Preprint, arXiv:2503.10150. find
  9. Hao Liu, Zhengren Wang, Xi Chen, Zhiyu Li, Feiyu Xiong, Qinhan Yu, and Wentao Zhang. 2025. HopRAG: Multi-Hop Reasoning for Logic-Aware Retrieval-Augmented Generation. Preprint, arXiv:2502.12442. find
  10. Vaibhav Mavi, Anubhav Jangra, and Adam Jatowt. 2024. Multi-hop question answering. Preprint, arXiv:2204.09140. — [arXiv:2204.09140](https://arxiv.org/abs/2204.09140) find
  11. Tyler McDonald and Ali Emami. 2024. Trace-of-thought prompting: Investigating prompt-based knowledge distillation through question decomposition. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 4: Student Research Workshop), pages 293–306, Bangkok,… find
  12. Sewon Min, Eric Wallace, Sameer Singh, Matt Gardner, Hannaneh Hajishirzi, and Luke Zettlemoyer. 2019. Compositional questions do not necessitate multi-hop reasoning. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4249–4257, Florence, Italy. Associat…
  13. Ofir Press, Muru Zhang, Sewon Min, Ludwig Schmidt, Noah Smith, and Mike Lewis. 2023. Measuring and Narrowing the Compositionality Gap in Language Models. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 5687–5711, Singapore. Association for Computational Linguistics. find
  14. Parth Sarthi, Salman Abdullah, Aditi Tuli, Shubh Khanna, Anna Goldie, and Christopher D. Manning. 2024. Raptor: Recursive abstractive processing for tree-organized retrieval. Preprint, arXiv:2401.18059.
  15. Hendrik Schuff, Heike Adel, and Ngoc Thang Vu. 2020. F1 is Not Enough! Models and Evaluation Towards User-Centered Explainable Question Answering. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 7076–7095, Online. Association for Computational… find
  16. Duolin Sun, Dan Yang, Yue Shen, Yihan Jiao, Zhehao Tan, Jie Feng, Lianzhen Zhong, Jian Wang, Peng Wei, and Jinjie Gu. 2025. Hanrag: Heuristic accurate noise-resistant retrieval-augmented generation for multi-hop question answering. Preprint, arXiv:2509.09713. — [arXiv:2509.09713](https://arxiv.org/abs/2509.09713) find
  17. Harsh Trivedi, Niranjan Balasubramanian, Tushar Khot, and Ashish Sabharwal. 2022. MuSiQue: Multi-hop Questions via Single-hop Question Composition. Transactions of the Association for Computational Linguistics, 10:539–554. find
  18. Harsh Trivedi, Niranjan Balasubramanian, Tushar Khot, and Ashish Sabharwal. 2023. Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questions. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), p…
  19. Yu Wang, Nedim Lipka, Ryan A. Rossi, Alexa Siu, Ruiyi Zhang, and Tyler Derr. 2023. Knowledge graph prompting for multi-document question answering. Preprint, arXiv:2308.11730. — [arXiv:2308.11730](https://arxiv.org/abs/2308.11730) find
  20. Tianyang Xu, Haojie Zheng, Chengze Li, Haoxiang Chen, Yixin Liu, Ruoxi Chen, and Lichao Sun. 2025. NodeRAG: Structuring Graph-based RAG with Heterogeneous Nodes. Preprint, arXiv:2504.11544. — [arXiv:2504.11544](https://arxiv.org/abs/2504.11544) find
  21. Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William Cohen, Ruslan Salakhutdinov, and Christopher D. Manning. 2018. HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 2369…
  22. Yaoze Zhang, Rong Wu, Pinlong Cai, Xiaoman Wang, Guohang Yan, Song Mao, Ding Wang, and Botian Shi. 2025. Leanrag: Knowledge-graph-based generation with semantic aggregation and hierarchical retrieval. Preprint, arXiv:2508.10391. — [arXiv:2508.10391](https://arxiv.org/abs/2508.10391) find
  23. Denny Zhou, Nathanael Schärli, Le Hou, Jason Wei, Nathan Scales, Xuezhi Wang, Dale Schuurmans, Claire Cui, Olivier Bousquet, Quoc Le, and Ed Chi. 2023. Least-to-most prompting enables complex reasoning in large language models. Preprint, arXiv:2205.10625.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/luu-2026-higraagent-dual-agent-adaptive-reasoning/