CYFRAGOVPL/PLLuM-4B-instruct-2512
The CYFRAGOVPL/PLLuM-4B-instruct-2512 is a 4 billion parameter instruction-tuned large language model, part of the PLLuM family developed by the PLLuM consortium and continued by HIVE AI. Specialized in Polish, it incorporates English data for generalization and is fine-tuned with extensive Polish instruction and preference datasets. This model excels in generating contextually coherent text for general language tasks and is particularly optimized for applications within Polish public administration.
Loading preview...
PLLuM-4B-instruct-2512: A Polish-Centric Instruction-Tuned LLM
The PLLuM-4B-instruct-2512 is a 4 billion parameter instruction-tuned model from the PLLuM family, developed by the PLLuM consortium and later by HIVE AI. It is built upon the gemma-3-4b-pt base model and is specifically designed for the Polish language, with additional English data for broader generalization. The model's development emphasizes high-quality data collection, including extensive Polish and English text corpora, rigorous cleaning, and deduplication.
Key Capabilities and Training Highlights
- Specialized Polish Instruction Tuning: Features the largest Polish collection of manually created "organic instructions" (approximately 70k), alongside 33k programmatically derived instructions and 45k synthetic, context-aware instructions. This unique dataset focuses on subtle aspects of supervised fine-tuning and mitigates negative linguistic transfer.
- Polish Preference Corpus: Utilizes the first Polish-language preference corpus (~60k manually annotated pairs) to enhance correctness, balance, and safety, particularly for sensitive topics.
- Optimized for Public Administration: Achieves top scores on custom benchmarks relevant to Polish public administration tasks and state-of-the-art results in broader Polish-language tasks.
- RAG-Ready: Post-training included specific optimization for Retrieval-Augmented Generation (RAG) scenarios, with a defined prompt format for document-based question answering.
Good for
- General Language Tasks: Text generation, summarization, extraction, and question answering in Polish.
- Domain-Specific Assistants: Especially effective for applications within Polish public administration, legal, or bureaucratic contexts requiring domain-aware retrieval.
- Research & Development: Serving as a foundational component for AI applications demanding strong Polish language command.