farbodtavakkoli/OTel-LLM-12B-IT

OTel-LLM-12B-IT: Telecom-Specialized Language Model

OTel-LLM-12B-IT is a 12 billion parameter language model developed by farbodtavakkoli, fine-tuned from the google/gemma-3-12b-it base model. It is a core component of the OTel Family of Models, an open-source initiative focused on building AI solutions for the telecommunications industry. The model underwent full parameter fine-tuning using an extensive dataset curated by over 100 domain experts from institutions like Yale University, GSMA, NetoAI, Khalifa University, and The University of Leeds.

Key Capabilities

• Domain Specialization: Highly proficient in understanding and generating content related to telecommunications, trained on data including arXiv telecom papers, 3GPP standards, IETF RFCs, and O-RAN specifications.
• RAG Optimization: Designed to power end-to-end Retrieval-Augmented Generation (RAG) pipelines, working in conjunction with OTel Embedding and Reranker models to provide context-grounded responses.
• Abstention Training: Incorporates abstention training, enabling the model to decline answering when insufficient context is provided, thereby minimizing hallucination and ensuring factual accuracy.

Good For

• Telecommunications RAG Systems: Ideal for building robust RAG applications that require precise, domain-specific information retrieval and generation within the telecom sector.
• Technical Documentation Analysis: Analyzing and extracting insights from complex telecom specifications, standards, and whitepapers.
• Context-Grounded Question Answering: Providing accurate answers to telecom-related queries, relying heavily on provided context rather than open-ended generation.