farbodtavakkoli/OTel-LLM-27B-IT
OTel-LLM-27B-IT by farbodtavakkoli is a 27 billion parameter, English-only language model fine-tuned on OTel telecommunications data. Based on google/gemma-3-27b-it, it is specifically optimized for context-grounded telecom answer generation within Retrieval-Augmented Generation (RAG) pipelines. This model demonstrates improved correctness in telecom-specific tasks, achieving a +3.3 percentage point increase over its base checkpoint in LLM-as-judge correctness on held-out OTel evaluation data.
Loading preview...
OTel-LLM-27B-IT: A Specialized Telecom Language Model
OTel-LLM-27B-IT is a 27 billion parameter language model developed by farbodtavakkoli, specifically fine-tuned on a comprehensive dataset of telecommunications data (OTel-LLM). This model is part of the broader OTel Family of Models, an initiative dedicated to providing open-source AI resources for the global telecom sector. It is built upon the google/gemma-3-27b-it base model and has undergone full-parameter post-training.
Key Capabilities
- Context-Grounded Telecom QA: Optimized for generating answers to telecom-related questions when provided with retrieved context, making it ideal for Retrieval-Augmented Generation (RAG) pipelines.
- Improved Correctness: Demonstrates a significant improvement in context-grounded correctness, achieving a +3.3 percentage point increase over its base model on OTel-LLM held-out evaluation data, as assessed by LLM-as-judge metrics.
- Domain-Specific Training: Trained on a curated dataset of 1.1 million raw telecom training points, filtered to 326,767 high-confidence examples, contributed by institutions like Yale University, GSMA, and Purdue University.
Good For
- Telecom RAG Applications: Best suited for applications requiring precise, context-grounded answers within the telecommunications domain.
- Specialized Telecom Research: Useful for researchers and developers working on AI solutions for 3GPP standards, O-RAN specifications, and other telecom-specific documentation.
- English-Only Telecom Tasks: Designed for English language processing in telecom contexts, where domain-specific accuracy is paramount.