farbodtavakkoli/OTel-LLM-32B-IT

OTel-LLM-32B-IT: A Specialized Telecom Language Model

OTel-LLM-32B-IT is a 32 billion parameter language model developed by farbodtavakkoli, specifically fine-tuned for the telecommunications sector. Built upon the allenai/OLMo-3-32B base model, it undergoes full-parameter post-training using the specialized OTel-LLM dataset, which comprises telecom-focused data curated by over 100 domain experts.

Key Capabilities and Differentiators

• Domain-Specific Expertise: Optimized for telecommunications, providing accurate, context-grounded answers within this specialized field.
• Enhanced Correctness: Demonstrates a significant improvement in context-grounded correctness, with a delta of +4.5 percentage points (from 66.8% to 71.3%) over its base model on held-out OTel evaluation data.
• RAG-Optimized: Primarily intended for use in Retrieval-Augmented Generation (RAG) pipelines, where it generates answers based on provided telecom context.
• Comprehensive Training Data: Trained on a filtered corpus of 326,767 high-confidence examples from diverse sources including arXiv telecom papers, 3GPP standards, GSMA documents, IETF RFCs, and O-RAN specifications.

Intended Use Cases

• Context-Grounded Telecom QA: Ideal for applications requiring precise answers to telecommunications questions when relevant context is provided.
• RAG Systems: Designed to be a core component in RAG architectures for telecom-specific information retrieval and synthesis.

Limitations

It is crucial to note that OTel-LLM-32B-IT is domain-specific and not a general-purpose language model. It is currently English-only and text-centric. Performance metrics are based on held-out OTel evaluation partitions, and generated content should always be verified for critical applications.