farbodtavakkoli/OTel-LLM-0.6B-IT

OTel-LLM-0.6B-IT: Telecom-Specialized Language Model

OTel-LLM-0.6B-IT is a 0.6 billion parameter language model, fine-tuned from the Qwen/Qwen3-0.6B base model, specifically for the telecommunications domain. It is part of the OTel Family of Models, an open-source initiative by Farbod Tavakkoli aimed at developing industry-standard AI for the global telecom sector.

Key Capabilities and Training

• Domain Specialization: Fully fine-tuned on a vast dataset curated by over 100 domain experts from institutions like Yale University, GSMA, NetoAI, Khalifa University, and the University of Leeds.
• Comprehensive Training Data: Includes arXiv telecom papers, 3GPP standards, telecom Wikipedia, Common Crawl, GSMA Permanent Reference Documents, IETF RFC series, industry whitepapers, and O-RAN specifications.
• RAG Pipeline Integration: Designed to power end-to-end Retrieval-Augmented Generation (RAG) pipelines for telecommunications, working in conjunction with OTel Embedding and Reranker models.
• Abstention Training: Incorporates abstention training, meaning the model will decline to answer if it does not receive sufficient context, thereby preventing hallucinations and ensuring context-grounded generation.

Intended Use Cases

• Telecom Information Retrieval: Generating accurate responses based on retrieved telecom specifications, standards, and documentation.
• Specialized Q&A: Answering questions within the telecommunications domain where factual accuracy and grounding in provided context are critical.
• RAG System Component: Serving as the generative component within a larger RAG system for telecom-specific applications.