farbodtavakkoli/OTel-LLM-27B-IT
OTel-LLM-27B-IT by farbodtavakkoli is a 27 billion parameter language model, fine-tuned from google/gemma-3-27b-it on specialized OTel telecommunications data. This model is specifically designed for context-grounded answer generation within Retrieval-Augmented Generation (RAG) pipelines for the telecommunications sector. It demonstrates improved correctness in telecom-specific contexts, outperforming its base model by +3.7 percentage points.
Loading preview...
OTel-LLM-27B-IT: Telecom-Specific Language Model
OTel-LLM-27B-IT is a 27 billion parameter language model developed by farbodtavakkoli, derived from google/gemma-3-27b-it. It has undergone full-parameter fine-tuning using the proprietary OTel-LLM dataset, which comprises over 326,000 high-confidence telecom-focused training points curated by domain experts. This model is part of the broader OTel Family of Models, an open-source initiative aimed at providing AI resources for the global telecommunications industry.
Key Capabilities and Differentiators
- Domain-Specific Expertise: Optimized for telecommunications data, including arXiv papers, 3GPP standards, GSMA documents, IETF RFCs, and O-RAN specifications.
- Enhanced Context-Grounded Correctness: Achieves an 88.2% correctness rate in LLM-as-judge evaluations on held-out OTel data, a +3.7 percentage point improvement over its base model.
- RAG Pipeline Integration: Specifically designed for generating answers grounded in retrieved telecom contexts.
- Apache 2.0 License: Released under an open-source license, promoting community use and development.
Intended Use Cases
- Context-Grounded Telecom QA: Ideal for Retrieval-Augmented Generation (RAG) systems requiring precise answers based on provided telecom-specific contexts.
- Research and Development: Useful for projects focused on AI applications within the telecommunications sector.
Limitations
- Domain-Specific: Not intended for general-purpose language tasks outside of telecommunications.
- English-Only: The current release is limited to the English language.
- Context-Dependent: Performance is optimized for scenarios where relevant context is provided; not designed for unrestricted, context-free question answering.