farbodtavakkoli/OTel-LLM-4B-IT

OTel-LLM-4B-IT: A Specialized Telecom Language Model

OTel-LLM-4B-IT is a 4.3 billion parameter language model developed by farbodtavakkoli, specifically fine-tuned for the telecommunications domain. Part of the OTel Family of Models, this model is built upon google/gemma-3-4b-it and has undergone full-parameter post-training using the OTel-LLM dataset.

Key Capabilities & Differentiators

• Domain-Specific Expertise: Optimized for context-grounded answer generation within telecommunications, leveraging a dataset curated by over 100 domain experts.
• Enhanced Correctness: Achieves a +6.5 percentage point improvement in LLM-as-judge correctness (from 66.2% to 72.7%) over its base model on held-out OTel evaluation data.
• RAG-Optimized: Primarily designed for Retrieval-Augmented Generation (RAG) pipelines, where it processes retrieved telecom context to generate precise answers.
• Comprehensive Training Data: Trained on a filtered corpus of 326,767 high-confidence examples derived from sources like arXiv telecom papers, 3GPP standards, GSMA documents, and O-RAN specifications.

Intended Use & Limitations

This model is intended for context-grounded telecom answer generation and should be used with retrieved telecom context. It is not optimized for unrestricted, context-free general question answering. Users should verify generated content, especially for operational, customer-facing, or regulatory applications. The model is currently English-only and text-centric, with performance evaluated on OTel-specific partitions rather than fully independent external benchmarks.