An End-to-End Ukrainian RAG for Local Deployment. Optimized Hybrid Search and Lightweight Generation

2026-04-23

AI Summary

This work proposes an end-to-end localized retrieval-augmented generation (RAG) system designed for efficient, accurate, and verifiable Ukrainian-language document question answering on resource-constrained devices. The system employs a Ukrainian-specific two-stage hybrid retrieval pipeline, integrates a lightweight language model fine-tuned on synthetically generated data, and leverages model compression techniques to operate effectively under stringent computational constraints. This approach significantly enhances both answer accuracy and verifiability while maintaining deployment efficiency. Evaluated in the UNLP 2026 shared task, the method achieved second place, demonstrating a strong balance between response quality and computational feasibility, and offering a practical solution for localized intelligent question answering in low-resource languages.

Technology Category

Natural Language Processing: Question AnsweringMachine Learning: Large Multimodal Models (LMMs)Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAGUser Modeling, Personalization and Recommendation: Fairness-aware retrieval and rankingWeb Mining and Content Analysis: Large pretrained models with web data

Abstract

This paper presents a highly efficient Retrieval-Augmented Generation (RAG) system built specifically for Ukrainian document question answering, which achieved 2nd place in the UNLP 2026 Shared Task. Our solution features a custom two-stage search pipeline that retrieves relevant document pages, paired with a specialized Ukrainian language model fine-tuned on synthetic data to generate accurate, grounded answers. Finally, we compress the model for lightweight deployment. Evaluated under strict computational limits, our architecture demonstrates that high-quality, verifiable AI question answering can be achieved locally on resource-constrained hardware without sacrificing accuracy.

Problem

• Retrieval-Augmented Generation
• Ukrainian language
• local deployment
• question answering
• resource-constrained hardware

Innovation

• Retrieval-Augmented Generation
• Hybrid Search
• Ukrainian Language Model
• Synthetic Data Fine-tuning
• Lightweight Deployment