Elamur Research⌕
☰
home / papers / 2026 / AMATA: Adaptive Multi-Agent Trajectory Alignment for Knowledge-Intensive Question Answering

AMATA: Adaptive Multi-Agent Trajectory Alignment for Knowledge-Intensive Question Answering

Paper · 2026 · Taolin Zhang, Dongyang Li, Chen Chen et al.

Summary

LLMs answering knowledge-intensive questions often hallucinate and struggle with long-tail facts. Existing multi-agent QA systems are trained in one of three ways: module by module, end to end, or in a global-local scheme. None of these adapts which agents run for a given question or models how agents depend on each other. AMATA addresses this with six specialized agents: intent reconstruction, retrieval, filtering, locating, response generation and verification. It treats their tool-using collaboration as a trajectory preference alignment task, which has two parts:

  • Intra-Trajectory Preference Learning attaches per-question utility scores for each agent to the fine-tuning input, so the system learns which agents matter for a given question.
  • Inter-Agent Dependency Learning uses a dependency-aware DPO (DA-DPO). It ranks sampled trajectories by a joint score that rewards coordinated agent chains.

Tested with 7B backbones on five QA benchmarks, AMATA beats plain baselines, knowledge-augmented methods and LLM trajectory systems. It also uses far fewer tokens than RL-style approaches. This makes adaptive, citation-capable multi-agent RAG cheaper to run.

Key points

  • The six agents are delimited by special head and end tokens, for example ⟨Retriever⟩ … ⟨/eor⟩. The knowledge agents (Retriever, Filter, Locator) are strongly interdependent, so triggering one implies running the others.
  • Benchmarks are HealthQA, ARC-Choice, PopQA, SQuAD 1.1 and ASQA. AMATA improves accuracy by +4.02% on average across them.
  • Token consumption is about 70% lower than strong baselines; RL-based methods such as GiGPO need many rollouts. Token use is comparable to SMART, but accuracy is +4.87% higher.
  • Base models are Llama2-Chat-7B and Qwen2.5-Instruct-7B. DA-DPO uses M = N = 10 winning and losing samples, with Top-K = 5 for ranking the winners.
  • Limitations:
    • Experiments use only 7B models.
    • On long trajectories (many agents, as in GUI tasks), RL methods catch up, while the preference-based methods accumulate errors.

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

🏢Alibaba GroupHangzhou · China201 works · 17,172 citations🎓East China Normal UniversityShanghai · China69 works · 2,490 citations🎓Hefei University of TechnologyHefei · China17 works · 1,702 citations🎓Shanghai University of Electric PowerYangpu · China5 works · 13 citations🎓Guangdong University Of Finances and EconomicsGuangzhou · China2 works

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksComputer Vision and Pattern Recognition · 3,486 works

🧭 Topics 1Research subjects and languages the work is about.Question answering · 487 works

Cite

Cite as Taolin Zhang, Dongyang Li, Chen Chen, Qizhou Chen, Jiuheng Wan, Xiaofeng He, Chengyu Wang and Richang Hong. 2026. AMATA: Adaptive Multi-Agent Trajectory Alignment for Knowledge-Intensive Question Answering. 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.359

@inproceedings{zhang2026amata,
  title = {{AMATA: Adaptive Multi-Agent Trajectory Alignment for Knowledge-Intensive Question Answering}},
  author = {Taolin Zhang, Dongyang Li, Chen Chen, Qizhou Chen, Jiuheng Wan, Xiaofeng He, Chengyu Wang and Richang Hong},
  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.359},
  url = {https://doi.org/10.18653/v1/2026.acl-long.359}
}

References

67 references, 24 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. OpenAI. 2023. Gpt-4 technical report. In OpenAI.
  2. Mubashara Akhtar, Oana Cocarascu, and Elena Simperl. 2022. PubHealthTab: A public health table-based dataset for evidence-based fact checking. In NAACL. find
  3. Akari Asai, Zeqiu Wu, Yizhong Wang, Avirup Sil, and Hannaneh Hajishirzi. 2024. Self-rag: Learning to retrieve, generate, and critique through self-reflection. In ICLR.
  4. Ralph Allan Bradley and Milton E. Terry. 1952. Rank analysis of incomplete block designs: I. the method of paired comparisons. Biometrika, 39(3/4):324– 345. find
  5. Baian Chen, Chang Shu, Ehsan Shareghi, Nigel Collier, Karthik Narasimhan, and Shunyu Yao. 2023. Fireact: Toward language agent fine-tuning. CoRR, abs/2310.05915.
  6. Jiaqi Chen, Bingqian Lin, Ran Xu, Zhenhua Chai, Xiaodan Liang, and Kwan-Yee Kenneth Wong. 2024. Mapgpt: Map-guided prompting with adaptive path planning for vision-and-language navigation. In ACL, pages 9796–9810. find
  7. Mingyue Cheng, Yucong Luo, Jie Ouyang, Qi Liu, Huijie Liu, Li Li, Shuo Yu, Bohou Zhang, Jiawei Cao, Jie Ma, Daoyu Wang, and Enhong Chen. 2025. A survey on knowledge-oriented retrieval-augmented generation. CoRR, abs/2503.10677. — [arXiv:2503.10677](https://arxiv.org/abs/2503.10677) find
  8. Peter Clark, Isaac Cowhey, Oren Etzioni, Tushar Khot, Ashish Sabharwal, Carissa Schoenick, and Oyvind Tafjord. 2018. Think you have solved question answering? try arc, the AI2 reasoning challenge. CoRR, abs/1803.05457.
  9. Huifang Du, Shuqin Li, Minghao Wu, Xuejing Feng, Yuan-Fang Li, and Haofen Wang. 2024. Rewarding what matters: Step-by-step reinforcement learning for task-oriented dialogue. In EMNLP, pages 8030– 8046. find
  10. Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, Anirudh Goyal, Anthony Hartshorn, Aobo Yang, Archi Mitra, Archie Sravankumar, Artem Korenev, Arthur Hinsvark, Arun Rao, Aston Zhang, and 82 others. 2024… — The Llama 3 Herd of Models
  11. Wenqi Fan, Yujuan Ding, Liangbo Ning, Shijie Wang, Hengyun Li, Dawei Yin, Tat-Seng Chua, and Qing Li. 2024. A survey on RAG meeting llms: Towards retrieval-augmented large language models. In KDD, pages 6491–6501.
  12. Feiteng Fang, Yuelin Bai, Shiwen Ni, Min Yang, Xiaojun Chen, and Ruifeng Xu. 2024. Enhancing noise robustness of retrieval-augmented language models with adaptive adversarial training. In ACL, pages 10028–10039.
  13. Lang Feng, Zhenghai Xue, Tingcong Liu, and Bo An. 2025. Group-in-group policy optimization for LLM agent training. CoRR, abs/2505.10978.
  14. Pengyu Gao, Jinming Zhao, Xinyue Chen, and Yilin Long. 2025. An efficient context-dependent memory framework for llm-centric agents. In NAACL, pages 1055–1069. find
  15. Tianyu Gao, Howard Yen, Jiatong Yu, and Danqi Chen. 2023. Enabling large language models to generate text with citations. In EMNLP, pages 6465–6488.
  16. Xinming Hou, Mingming Yang, Wenxiang Jiao, Xing Wang, Zhaopeng Tu, and Wayne Xin Zhao. 2024. Coact: A global-local hierarchy for autonomous agent collaboration. CoRR, abs/2406.13381. — [arXiv:2406.13381](https://arxiv.org/abs/2406.13381) find
  17. Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2022. Lora: Low-rank adaptation of large language models. In ICLR.
  18. Yuchen Hu, Chen Chen, Chao-Han Huck Yang, Ruizhe Li, Dong Zhang, Zhehuai Chen, and Engsiong Chng. 2024. Gentranslate: Large language models are generative multilingual speech and machine translators. In ACL, pages 74–90. find
  19. Lei Huang, Weijiang Yu, Weitao Ma, Weihong Zhong, Zhangyin Feng, Haotian Wang, Qianglong Chen, Weihua Peng, Xiaocheng Feng, Bing Qin, and Ting Liu. 2025. A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions. ACM Trans. Inf. Syst., 43(2):42:1– 42:55…
  20. Yuhan Ji and Song Gao. 2024. Evaluating the effectiveness of large language models in representing and understanding movement trajectories. CoRR, abs/2409.00335. — [arXiv:2409.00335](https://arxiv.org/abs/2409.00335) find
  21. Ziwei Ji, Tiezheng Yu, Yan Xu, Nayeon Lee, Etsuko Ishii, and Pascale Fung. 2023. Towards mitigating LLM hallucination via self reflection. In EMNLP, pages 1827–1843.
  22. Albert Q. Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de Las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lucile Saulnier, Lélio Renard Lavaud, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix,… find
  23. Vladimir Karpukhin, Barlas Oguz, Sewon Min, Patrick S. H. Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih. 2020. Dense passage retrieval for open-domain question answering. In EMNLP, pages 6769–6781.
  24. Tushar Khot, Harsh Trivedi, Matthew Finlayson, Yao Fu, Kyle Richardson, Peter Clark, and Ashish Sabharwal. 2023. Decomposed prompting: A modular approach for solving complex tasks. In ICLR.
  25. Dorde Klisura, Astrid R. Bernaga Torres, Anna Karen Gárate-Escamilla, Rajesh Roshan Biswal, Ke Yang, Hilal Pataci, and Anthony Rios. 2025. A multi-agent framework for mitigating dialect biases in privacy policy question-answering systems. CoRR, abs/2506.02998. find
  26. Bevan Koopman, Ahmed Mourad, Hang Li, Anton van der Vegt, Shengyao Zhuang, Simon Gibson, Yash Dang, David Lawrence, and Guido Zuccon. 2024. Agask: an agent to help answer farmer’s questions from scientific documents. Int. J. Digit. Libr., 25(4):569–584. find
  27. Teyun Kwon, Norman Di Palo, and Edward Johns. 2024. Language models as zero-shot trajectory generators. IEEE Robotics Autom. Lett., 9(7):6728–6735. find
  28. Dongyang Li, Junbing Yan, Taolin Zhang, Chengyu Wang, Xiaofeng He, Longtao Huang, Hui Xue, and Jun Huang. 2024. On the role of long-tail knowledge in retrieval augmented large language models. In ACL, pages 120–126. Association for Computational Linguistics. find
  29. Xinzhe Li. 2025. A review of prominent paradigms for llm-based agents: Tool use, planning (including rag), and feedback learning. In COLING, pages 9760– 9779. find
  30. Xi Victoria Lin, Xilun Chen, Mingda Chen, Weijia Shi, Maria Lomeli, Richard James, Pedro Rodriguez, Jacob Kahn, Gergely Szilvasy, Mike Lewis, Luke Zettlemoyer, and Wen-tau Yih. 2024. RA-DIT: retrieval-augmented dual instruction tuning. In ICLR.
  31. Yuxing Long, Binyuan Hui, Fulong Ye, Yanyang Li, Zhuoxin Han, Caixia Yuan, Yongbin Li, and Xiaojie Wang. 2023. SPRING: situated conversation agent pretrained with multimodal questions from incremental layout graph. In AAAI, pages 13309–13317. find
  32. Alex Mallen, Akari Asai, Victor Zhong, Rajarshi Das, Hannaneh Hajishirzi, and Daniel Khashabi. 2022. When not to trust language models: Investigating effectiveness and limitations of parametric and non-parametric memories. CoRR, abs/2212.10511. find
  33. Silin Meng, Yiwei Wang, Cheng-Fu Yang, Nanyun Peng, and Kai-Wei Chang. 2024. Llm-a*: Large language model enhanced incremental heuristic search on path planning. In EMNLP, pages 1087–1102. Association for Computational Linguistics. find
  34. R. L. Plackett. 1975. The analysis of permutations. Journal of the Royal Statistical Society. Series C (Applied Statistics), 24(2):193–202. find
  35. Chen Qian, Zihao Xie, Yifei Wang, Wei Liu, Yufan Dang, Zhuoyun Du, Weize Chen, Cheng Yang, Zhiyuan Liu, and Maosong Sun. 2025. Scaling large-language-model-based multi-agent collaboration. In ICLR.
  36. Shuofei Qiao, Zhisong Qiu, Baochang Ren, Xiaobin Wang, Xiangyuan Ru, Ningyu Zhang, Xiang Chen, Yong Jiang, Pengjun Xie, Fei Huang, and Huajun Chen. 2025. Agentic knowledgeable self-awareness. In ACL, pages 12601–12625. find
  37. Qwen Team. 2024. Qwen2.5: A party of foundation models. find
  38. Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D. Manning, Stefano Ermon, and Chelsea Finn. 2023. Direct preference optimization: Your language model is secretly a reward model. In NeurIPS.
  39. Pranav Rajpurkar, Jian Zhang, Konstantin Lopyrev, and Percy Liang. 2016. Squad: 100, 000+ questions for machine comprehension of text. In EMNLP, pages 2383–2392.
  40. Jeff Rasley, Samyam Rajbhandari, Olatunji Ruwase, and Yuxiong He. 2020. Deepspeed. In KDD. find
  41. Ohad Rubin, Jonathan Herzig, and Jonathan Berant. 2022. Learning to retrieve prompts for in-context learning. In NAACL, pages 2655–2671.
  42. Weijia Shi, Sewon Min, Michihiro Yasunaga, Minjoon Seo, Richard James, Mike Lewis, Luke Zettlemoyer, and Wen-tau Yih. 2024. REPLUG: retrieval-augmented black-box language models. In NAACL, pages 8371–8384.
  43. Aditi Singh, Abul Ehtesham, Saket Kumar, and Tala Talaei Khoei. 2025. Agentic retrieval-augmented generation: A survey on agentic RAG. CoRR, abs/2501.09136. — [arXiv:2501.09136](https://arxiv.org/abs/2501.09136) find
  44. Yifan Song, Da Yin, Xiang Yue, Jie Huang, Sujian Li, and Bill Yuchen Lin. 2024. Trial and error: Exploration-based trajectory optimization for LLM agents. CoRR, abs/2403.02502. — [arXiv:2403.02502](https://arxiv.org/abs/2403.02502) find
  45. Arjun Subramonian, Xingdi Yuan, Hal Daumé III, and Su Lin Blodgett. 2023. It takes two to tango: Navigating conceptualizations of NLP tasks and measurements of performance. In ACL, pages 3234–3279. find
  46. Chuanneng Sun, Songjun Huang, and Dario Pompili. 2024. Llm-based multi-agent reinforcement learning: Current and future directions. CoRR, abs/2405.11106. — [arXiv:2405.11106](https://arxiv.org/abs/2405.11106) find
  47. Xiaolong Tang, Meina Kan, Shiguang Shan, and Xilin Chen. 2025. Plan-r1: Safe and feasible trajectory planning as language modeling. CoRR, abs/2505.17659. — [arXiv:2505.17659](https://arxiv.org/abs/2505.17659) find
  48. Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, Aurélien Rodriguez, Armand Joulin, Edouard Grave, and Guillaume Lample. 2023. Llama: Open and efficient foundation language models. CoRR, ab…
  49. Hanlin Wang, Chak Tou Leong, Jiashuo Wang, Jian Wang, and Wenjie Li. 2025a. SPA-RL: reinforcing LLM agents via stepwise progress attribution. CoRR, abs/2505.20732. — [arXiv:2505.20732](https://arxiv.org/abs/2505.20732) find
  50. Yuhao Wang, Ruiyang Ren, Yucheng Wang, Wayne Xin Zhao, Jing Liu, Hua Wu, and Haifeng Wang. 2025b. Unveiling knowledge utilization mechanisms in llm-based retrieval-augmented generation. CoRR, abs/2505.11995. — [arXiv:2505.11995](https://arxiv.org/abs/2505.11995) find
  51. Sangmin Woo, Kang Zhou, Yun Zhou, Shuai Wang, Sheng Guan, Haibo Ding, and Lin Lee Cheong. 2025. Black-box visual prompt engineering for mitigating object hallucination in large vision language models. In NAACL, pages 529–538. find
  52. Weimin Xiong, Yifan Song, Xiutian Zhao, Wenhao Wu, Xun Wang, Ke Wang, Cheng Li, Wei Peng, and Sujian Li. 2024. Watch every step! LLM agent learning via iterative step-level process refinement. In EMNLP, pages 1556–1572. find
  53. Heng-Da Xu, Xian-Ling Mao, Puhai Yang, Fanshu Sun, and Heyan Huang. 2024a. Rethinking task-oriented dialogue systems: From complex modularity to zero-shot autonomous agent. In ACL, pages 2748–2763. find
  54. Shicheng Xu, Liang Pang, Mo Yu, Fandong Meng, Huawei Shen, Xueqi Cheng, and Jie Zhou. 2024b. Unsupervised information refinement training of large language models for retrieval-augmented generation. In ACL, pages 133–145. find
  55. Cheng Yang, Chufan Shi, Siheng Li, Bo Shui, Yujiu Yang, and Wai Lam. 2025. LLM2: let large language models harness system 2 reasoning. In NAACL, pages 168–177. find
  56. Shunyu Yao, Howard Chen, John Yang, and Karthik Narasimhan. 2022. Webshop: Towards scalable real-world web interaction with grounded language agents. In NeurIPS.
  57. Fuda Ye, Shuangyin Li, Yongqi Zhang, and Lei Chen. 2024. R 2 ag: Incorporating retrieval information into retrieval augmented generation. In EMNLP, pages 11584–11596. find
  58. Shengbin Yue, Siyuan Wang, Wei Chen, Xuanjing Huang, and Zhongyu Wei. 2025. Synergistic multi-agent framework with trajectory learning for knowledge-intensive tasks. In AAAI, pages 25796– 25804. find
  59. Zhenrui Yue, Huimin Zeng, Lanyu Shang, Yifan Liu, Yang Zhang, and Dong Wang. 2024. Retrieval augmented fact verification by synthesizing contrastive arguments. In ACL, pages 10331–10343. find
  60. Aohan Zeng, Mingdao Liu, Rui Lu, Bowen Wang, Xiao Liu, Yuxiao Dong, and Jie Tang. 2024. Agenttuning: Enabling generalized agent abilities for llms. In ACL, pages 3053–3077.
  61. Guibin Zhang, Yanwei Yue, Zhixun Li, Sukwon Yun, Guancheng Wan, Kun Wang, Dawei Cheng, Jeffrey Xu Yu, and Tianlong Chen. 2025a. Cut the crap: An economical communication pipeline for llm-based multi-agent systems. In ICLR. find
  62. Shaokun Zhang, Ming Yin, Jieyu Zhang, Jiale Liu, Zhiguang Han, Jingyang Zhang, Beibin Li, Chi Wang, Huazheng Wang, Yiran Chen, and Qingyun Wu. 2025b. Which agent causes task failures and when? on automated failure attribution of LLM multi-agent systems. CoRR, abs/2505.00212. — [arXiv:2505.00212](https://arxiv.org/abs/2505.00212) find
  63. Taolin Zhang, Dongyang Li, Qizhou Chen, Chengyu Wang, Longtao Huang, Hui Xue, Xiaofeng He, and Jun Huang. 2024. R4 : Reinforced retriever-reorder-responder for retrieval-augmented large language models. In 27th European Conference on Artificial Intelligence, Frontiers in Artificial Intelligence and… find
  64. Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric P. Xing, Hao Zhang, Joseph E. Gonzalez, and Ion Stoica. 2023. Judging llm-as-a-judge with mt-bench and chatbot arena. In NeurIPS.
  65. Kun Zhu, Xiaocheng Feng, Xiyuan Du, Yuxuan Gu, Weijiang Yu, Haotian Wang, Qianglong Chen, Zheng Chu, Jingchang Chen, and Bing Qin. 2024. An information bottleneck perspective for effective noise filtering on retrieval-augmented generation. In ACL, pages 1044–1069. find
  66. Chang Zong, Yuchen Yan, Weiming Lu, Jian Shao, Yongfeng Huang, Heng Chang, and Yueting Zhuang. 2024. Triad: A framework leveraging a multi-role llm-based agent to solve knowledge base question answering. In EMNLP, pages 1698–1710. find
  67. Longwei Zou, Qingyang Wang, Han Zhao, Jiangangkong Jiangangkong, Yi Yang, and Yangdong Deng. 2024. CQIL: inference latency optimization with concurrent computation of quasi-independent layers. In ACL, pages 7293–7307. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zhang-2026-amata-adaptive-multi-agent-trajectory/