farbodtavakkoli/OTel-LLM-12B-Safety
The farbodtavakkoli/OTel-LLM-12B-Safety is a 12-billion parameter, safety-tuned telecom language model, full-parameter fine-tuned on OTel telecommunications data. Derived from google/gemma-3-12b-it, its primary differentiator is its specialized training to abstain from answering when retrieved context is insufficient or off-topic. This model is designed for telecom RAG pipelines where reliable refusal behavior is critical, ensuring responses are grounded in provided context.
Loading preview...
OTel-LLM-12B-Safety: Safety-Tuned Telecom LLM
OTel-LLM-12B-Safety is a 12-billion parameter language model developed by farbodtavakkoli, specifically fine-tuned for the telecommunications sector. Built upon the google/gemma-3-12b-it base model, it is a specialized variant within the OTel Family of Models, an open-source initiative focused on AI resources for global telecommunications.
Key Capabilities
- Contextual Abstention: Uniquely trained to refuse to answer when the retrieved context is insufficient or off-topic, enhancing reliability in RAG applications.
- Telecom Domain Expertise: Full-parameter fine-tuned on the OTel-Safety dataset, which comprises telecom-focused data curated by over 100 domain experts.
- Robust Training: Utilizes full-parameter post-training with a focus on abstention behavior, ensuring the model's responses are strictly grounded in relevant information.
Good For
- Telecom RAG Pipelines: Ideal for Retrieval Augmented Generation (RAG) systems in telecommunications where precise, context-bound answers are essential and abstention from ungrounded responses is desired.
- Safety-Critical Applications: Suitable for scenarios requiring a high degree of answer reliability and the ability to explicitly indicate when information is lacking.
- Reference for Further Tuning: Can serve as a strong checkpoint for additional safety tuning or specialized abstention-focused development within the telecom domain.