farbodtavakkoli/OTel-LLM-8.3B-Safety

OTel-LLM-8.3B-Safety: Telecom-Specialized Language Model

OTel-LLM-8.3B-Safety is an 8.3 billion parameter language model developed by farbodtavakkoli, specifically fine-tuned for the telecommunications sector. It is built on the EssentialAI/rnj-1-instruct base model and is part of the broader OTel Family of Models, an open-source initiative focused on AI for global telecom.

Key Capabilities & Training

• Domain-Specific Expertise: Trained extensively on telecom-focused data, including arXiv papers, 3GPP standards, GSMA documents, IETF RFCs, O-RAN specifications, and industry whitepapers, curated by over 100 domain experts from institutions like Yale University and the University of Leeds.
• RAG Pipeline Integration: Designed to function as the generative component within end-to-end Retrieval-Augmented Generation (RAG) pipelines for telecommunications. It complements OTel Embedding and Reranker models to provide accurate, context-grounded responses.
• Abstention Training: Features abstention training, meaning the model is optimized to decline to answer if it does not receive sufficient context, thereby minimizing hallucination and ensuring responses are grounded in provided information.

Intended Use Cases

• Context-Grounded Generation: Ideal for generating responses based on retrieved telecom specifications, standards, and documentation.
• Telecom RAG Systems: Suitable for deployment as the LLM component in RAG pipelines requiring high accuracy and domain specificity in the telecommunications field.
• Specialized Q&A: Can be used for question answering within the telecom domain where answers must be strictly derived from provided context.