SavedAdd your journal

Journals / International Journal of 3D Printing Technologies and Digital Industry / 2026 / Cilt: 10 Sayı: 2

ADAPTIVE TURKISH FEDERATED RAG ARCHITECTURE FOR LOW-RESOURCE SYSTEMS

Mehmet Taha Toy(ISPARTA UYGULAMALI BİLİMLER ÜNİVERSİTESİ)0009-0000-4220-6224
Ahmet Ali Süzen(Isparta Uygulamalı Bilimler Üniversitesi)0000-0002-5871-1652
Pages
300–317
Publication Date
2026-08-31
🏛️Bibliographic metadata and full text access for this article are provided via DergiPark.
DergiPark Record ↗
✨

AI Summary & Key Takeaways

AI Insight
⚡ TL;DR(Too Long; Didn't Read)

However, limitations such as encryption latency, erroneous information from untrusted sources, and the inadequacy of standard evaluation methods for Turkish also exist.

📌 3 Key Takeaways

🎯
Objective & Hypothesis:Today, Large Language Models (LLMs) perform many tasks in the field of natural language processing with high success, from text generation to translation, semantic analysis to code writing.
📊
Key Findings & Data:This study examines distributed learning approaches aimed at protecting data privacy and QLoRA-based fine-tuning strategies within a mathematical framework.
💡
Conclusion & Impact:However, limitations such as encryption latency, erroneous information from untrusted sources, and the inadequacy of standard evaluation methods for Turkish also exist.
🔬Design:Modeling & Simulation
📍Scope:Comprehensive Sample
⚙️Variables:Federated Learning · RAG · Turkish NLP

Abstract

Today, Large Language Models (LLMs) perform many tasks in the field of natural language processing with high success, from text generation to translation, semantic analysis to code writing. However, these models have some fundamental limitations that make their reliable use challenging. They can produce factual errors known in the literature as hallucinations and cannot directly access developments after their training period. They also sometimes reproduce biases present in the training data. The Re-trieval-Augmented Generation (RAG) approach aims to produce more up-to-date and verifiable outputs by dynamically feeding the model with external information sources, thereby reducing hallucination and temporal limitations. This study examines distributed learning approaches aimed at protecting data privacy and QLoRA-based fine-tuning strategies within a mathematical framework. It also addresses 4-bit NormalFloat (NF4) quantization techniques used to improve system efficiency. Furthermore, it de-tails morphology-aware hybrid access methods and adaptive routing mechanisms that can make deci-sions based on query complexity to achieve results more suitable for Turkish. The study demonstrates that combining RAG-based systems with federated learning and homomorphic encryption creates a secure, decentralized architecture, and that LLM can be efficiently run on low-resource systems using the NF4-QLoRA combination. However, limitations such as encryption latency, erroneous information from untrusted sources, and the inadequacy of standard evaluation methods for Turkish also exist.

Özet

Today, Large Language Models (LLMs) perform many tasks in the field of natural language processing with high success, from text generation to translation, semantic analysis to code writing. However, these models have some fundamental limitations that make their reliable use challenging. They can produce factual errors known in the literature as hallucinations and cannot directly access developments after their training period. They also sometimes reproduce biases present in the training data. The Re-trieval-Augmented Generation (RAG) approach aims to produce more up-to-date and verifiable outputs by dynamically feeding the model with external information sources, thereby reducing hallucination and temporal limitations. This study examines distributed learning approaches aimed at protecting data privacy and QLoRA-based fine-tuning strategies within a mathematical framework. It also addresses 4-bit NormalFloat (NF4) quantization techniques used to improve system efficiency. Furthermore, it de-tails morphology-aware hybrid access methods and adaptive routing mechanisms that can make deci-sions based on query complexity to achieve results more suitable for Turkish. The study demonstrates that combining RAG-based systems with federated learning and homomorphic encryption creates a secure, decentralized architecture, and that LLM can be efficiently run on low-resource systems using the NF4-QLoRA combination. However, limitations such as encryption latency, erroneous information from untrusted sources, and the inadequacy of standard evaluation methods for Turkish also exist.

🔖

Cite This Article

Toy, M. T., Süzen, A. A. (2026). ADAPTIVE TURKISH FEDERATED RAG ARCHITECTURE FOR LOW-RESOURCE SYSTEMS. International Journal of 3D Printing Technologies and Digital Industry, Cilt: 10 Sayı: 2, 300–317. https://doi.org/10.46519/ij3dptdi.1895770

📄 Full Text PDF Viewer

📄

PDF görüntüleyici hazırlanıyor...

References (29)

  1. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł. and Po-losukhin, I., "Attention is all you need", Advanc-es in Neural Information Processing Systems, Vol. 30, 2017.
  2. Radford, A., Wu, J., Child, R., Luan, D., Amo-dei, D. and Sutskever, I., "Language models are unsupervised multitask learners", OpenAI Blog, Vol. 1, Issue 8, Page 9, 2019.
  3. Ji, Z., Lee, N., Frieske, R., Yu, T., Su, D., Xu, Y., Ishii, E., Bang, Y. J., Madotto, A. and Fung, P., "Survey of hallucination in natural language generation", ACM Computing Surveys, Vol. 55, Issue 12, Pages 1-38, 2023.
  4. Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Küttler, H., Lewis, M., Yih, W., Rocktäschel, T., Riedel, S. and Kiela, D., "Retrieval-augmented generation for knowledge-intensive nlp tasks", arXiv preprint arXiv:2005.11401, 2020.
  5. Yang, Q., Liu, Y., Chen, T. and Tong, Y., "Fed-erated machine learning: Concept and applica-tions", ACM Transactions on Intelligent Systems and Technology, Vol. 10, Issue 2, Pages 1-19, 2019.
  6. Eryiğit, G., Nivre, J. and Oflazer, K., "Depend-ency parsing of Turkish", Computational Lin-guistics, Vol. 34, Issue 3, Pages 357-389, 2008.
  7. Gao, Y., Xiong, Y., Gao, X., Jia, K., Pan, J., Bi, Y., Dai, Y., Sun, J. and Wang, H., "Retrieval-augmented generation for large language mod-els: A survey", arXiv preprint arXiv:2312.10997, 2024.
  8. Gumaan, E., "ExpertRAG: Efficient RAG with mixture of experts -- Optimizing context retriev-al for adaptive LLM responses", arXiv preprint arXiv:2504.08744, 2025.
  9. Jeong, S., Baek, J., Cho, S., Hwang, S. J. and Park, J. C., "Adaptive-RAG: Learning to adapt retrieval-augmented large language models through question complexity", arXiv preprint arXiv:2403.14403, 2024.
  10. Reimers, N. and Gurevych, I., "Sentence-BERT: Sentence embeddings using Siamese BERT-networks", arXiv preprint arXiv:1908.10084, 2019.
  11. Sennrich, R., Haddow, B. and Birch, A., "Neu-ral machine translation of rare words with sub-word units", arXiv preprint arXiv:1508.07909, 2016.
  12. Şahin, G. G. and Steedman, M., "Data aug-mentation via dependency tree morphing for low-resource languages", arXiv preprint arXiv:1903.09460, 2018.
  13. Eryiğit, G., "ITU Turkish NLP web service", Proceedings of the Demonstrations at the 14th Conference of the European Chapter of the As-sociation for Computational Linguistics, Pages 1-4, 2014.
  14. Akarsu, M., Karaman, R. K., & Mierbach, C., “From BM25 to Corrective RAG: Benchmarking Retrieval Strategies for Text-and-Table Docu-ments”. Retrieved from https://arxiv.org/abs/2604.01733, 2025.
  15. Taş, S., El Huseyni, M., Ezerceli, Ö., Bay-raktar, R. and Terzioğlu, F. B., "Turk-LettuceDetect: A Hallucination Detection Mod-els for Turkish RAG Applications", arXiv pre-print arXiv:2509.17671, 2025.
  16. Mao, Q., Yuan, Z., Yin, J., Jiang, W., Wang, L., Liang, J., Dong, S., Qian, S. and Zhao, D., "Privacy-Preserving Federated Embedding Learning for Localized Retrieval-Augmented Generation", arXiv preprint arXiv:2504.19101, 2025.
  17. Addison, P., Nguyen, M. H., Medan, T., Shah, J., Manzari, M. T., Shen, S., Hashemi, M., Gupta, U. and Reddi, V. J., "C-FedRAG: A Confidential Federated Retrieval-Augmented Generation System", arXiv preprint arXiv:2412.13163, 2024.
  18. Zhang, Y., Liu, Y., Tan, J. and Huang, J., "RAFFLE: Retrieval augmented few-shot learn-ing with language models for federated learn-ing", arXiv preprint arXiv:2307.09284, 2023.
  19. Gentry, C., "A fully homomorphic encryption scheme", Doctoral Dissertation, Stanford Uni-versity, Stanford, 2009.
  20. Korkmaz, A. and Rao, P., "A Selective Ho-momorphic Encryption Approach for Faster Privacy-Preserving Federated Learning", arXiv preprint arXiv:2501.12911, 2025.
  21. Hong, C., “Recent advances of privacy-preserving machine learning based on (Fully) Homomorphic Encryption.” Security and Safety, 4, 2024012,2025.
  22. Ao, W. et al., "AutoFHE: Automated Adaption of CNNs for Efficient Evaluation over FHE", IACR ePrint, 2023.
  23. Bonawitz, K., Ivanov, V., Kreuter, B., Mar-cedone, A., McMahan, H. B., Patel, S., Ramage, D., Segal, A. and Seth, K., "Practical secure ag-gregation for privacy-preserving machine learn-ing", Proceedings of the 2017 ACM SIGSAC Conference on Computer and Communications Security, Pages 1175-1191, 2017.
  24. Dwork, C. and Roth, A., "The algorithmic foundations of differential privacy", Founda-tions and Trends in Theoretical Computer Sci-ence, Vol. 9, Issue 3-4, Pages 211-407, 2014.
  25. Dettmers, T., Pagnoni, A., Holtzman, A. and Zettlemoyer, L., "QLoRA: Efficient finetuning of quantized LLMs", arXiv preprint arXiv:2305.14314, 2023.
  26. Mekala, A., Atmakuru, A., Song, Y., Karpins-ka, M., & Iyyer, M. “Does quantization affect models’ performance on long-context tasks?”, Retrieved from https://arxiv.org/abs/2505.20276, 2025.
  27. Hu, E. J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Wang, L. and Chen, W., "LoRA: Low-rank adaptation of large language models", arXiv preprint arXiv:2106.09685, 2025.
  28. Toy, M. T., & Süzen, A. A. “Advantages And Disadvantages of Retrieval Augment-Ed Genera-tion (Rag) In Generative AI”. International Jour-nal of 3D Printing Technologies and Digital Industry, Vol. 9, Issue 3, Pages 462-475, 2025.
  29. Mekala, A., Atmakuru, A., Song, Y., Karpins-ka, M., & Iyyer, M. “Does quantization affect models’ performance on long-context tasks?”, Retrieved from https://arxiv.org/abs/2505.20276, 2025.