TeaRAG: A Token-Efficient Agentic Retrieval-Augmented Generation Framework
ACM Transactions on Information Systems · 31.07.2026 · 44 (6) · 1-35
DOI: 10.1145/3818621 ↗Kurz gefasst
Fachartikel in der Zeitschrift ACM Transactions on Information Systems, online erschienen am 20.07.2026. Verfasst von Chao Zhang und 11 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 91 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Science and Technology Major Project, Anhui Natural Science Foundation, National Natural Science Foundation of China.
Abstract (Originalsprache)
Retrieval-Augmented Generation (RAG) utilizes external knowledge to augment Large Language Models’ (LLMs) reliability. For flexibility, agentic RAG employs autonomous, multi-round retrieval and reasoning to resolve queries. Although recent agentic RAG has improved via reinforcement learning, they often incur substantial token overhead from search and reasoning. This tradeoff prioritizes accuracy over efficiency. To address this issue, this work proposes TeaRAG, a T oken- e fficient a gentic RAG framework capable of compressing both retrieval content and reasoning steps. (1) First, the retrieved content is compressed by augmenting chunk-based semantic retrieval with a graph retrieval using concise triplets. A knowledge association graph is then built from semantic similarity and co-occurrence. Finally, Personalized PageRank is leveraged to highlight key knowledge within this graph, reducing the number of tokens per retrieval. (2) Besides, to reduce reasoning steps, Iterative Process-aware Direct Preference Optimization (IP-DPO) is proposed. Specifically, our reward function evaluates the knowledge sufficiency by a knowledge matching mechanism, while penalizing excessive reasoning steps. This design can produce high-quality preference-pair datasets, supporting iterative DPO to improve reasoning conciseness. Across six datasets, TeaRAG improves the average Exact Match by \(4\%\) and \(2\%\) while reducing output tokens by \(61\%\) and \(59\%\) on Llama3-8B-Instruct and Qwen2.5-14B-Instruct, respectively. Code is available at https://github.com/Applied-Machine-Learning-Lab/TeaRAG .
Förderung
- National Science and Technology Major Project
- Anhui Natural Science Foundation
- National Natural Science Foundation of China
- Hong Kong Research Grants Council
- Institute of Digital Medicine of City University of Hong Kong
- Huawei
- Tencent
- Kuaishou
Zitieren
APA
Zhang, C., Wang, Y., Xu, D., Zhang, H., Lyu, Y., Chen, Y., Liu, S., Xu, T., Zhao, X., Gao, Y., Hu, Y., & Chen, E. (2026). TeaRAG: A Token-Efficient Agentic Retrieval-Augmented Generation Framework. ACM Transactions on Information Systems, 44(6), 1-35. https://doi.org/10.1145/3818621
Vancouver
Zhang C, Wang Y, Xu D, Zhang H, Lyu Y, Chen Y, et al. TeaRAG: A Token-Efficient Agentic Retrieval-Augmented Generation Framework. ACM Transactions on Information Systems. 2026;44(6):1-35. doi:10.1145/3818621
BibTeX
@article{Zhang2026,
author = {Zhang, Chao and Wang, Yuhao and Xu, Derong and Zhang, Haoxin and Lyu, Yuanjie and Chen, Yuhao and Liu, Shuochen and Xu, Tong and Zhao, Xiangyu and Gao, Yan and Hu, Yao and Chen, Enhong},
title = {TeaRAG: A Token-Efficient Agentic Retrieval-Augmented Generation Framework},
journal = {ACM Transactions on Information Systems},
year = {2026},
volume = {44},
number = {6},
pages = {1--35},
doi = {10.1145/3818621}
}
EUVIMED