farbodtavakkoli/OTel-LLM-24B-IT
farbodtavakkoli/OTel-LLM-24B-IT is a 24 billion parameter language model developed by farbodtavakkoli, fine-tuned from LiquidAI/LFM2-24B-A2B. This model is specifically optimized for context-grounded answer generation within the telecommunications domain, utilizing the OTel telecommunications dataset. It demonstrates improved correctness in telecom-specific context-grounded tasks, showing a +4.5 percentage point increase over its base model. The primary use case is for Retrieval-Augmented Generation (RAG) pipelines requiring precise telecom-related responses.
Loading preview...
OTel-LLM-24B-IT: A Specialized Telecom Language Model
OTel-LLM-24B-IT is a 24 billion parameter language model developed by farbodtavakkoli, specifically fine-tuned for the telecommunications sector. Built upon the LiquidAI/LFM2-24B-A2B base model, it leverages the proprietary OTel telecommunications dataset, curated by over 100 domain experts, to enhance its domain-specific knowledge and performance.
Key Capabilities and Differentiators
- Domain-Specific Optimization: Full-parameter fine-tuned on a comprehensive OTel dataset, including arXiv telecom papers, 3GPP standards, GSMA documents, and O-RAN specifications.
- Enhanced Context-Grounded Correctness: Achieves a +4.5 percentage point improvement in LLM-as-judge correctness (79.5%) over its base model (75.0%) on held-out OTel evaluation partitions, specifically for context-grounded answer generation.
- RAG Pipeline Integration: Designed for use in Retrieval-Augmented Generation (RAG) pipelines, where it processes retrieved telecom context to generate accurate, grounded answers.
- Extensive Training Data: Trained on a filtered dataset of 326,767 high-confidence telecom examples, ensuring deep specialization.
Intended Use Cases
- Telecom RAG Systems: Ideal for applications requiring precise, context-grounded answers to telecom-related queries, such as technical support, documentation analysis, or network management.
- Specialized QA: Suitable for question-answering systems where the input includes relevant telecom context.
Limitations
- Domain-Specific: Not intended as a general-purpose language model; its performance is optimized for telecommunications.
- English-Only: The current release is limited to the English language.
- Context-Dependent: Performance is tied to the quality and sufficiency of the retrieved context; it is not optimized for unrestricted, context-free question answering.