90 lines
3.2 KiB
Markdown
90 lines
3.2 KiB
Markdown
---
|
|
license: apache-2.0
|
|
language:
|
|
- en
|
|
base_model:
|
|
- google/gemma-3-4b-it
|
|
tags:
|
|
- telecom
|
|
- telecommunications
|
|
- gsma
|
|
- fine-tuned
|
|
pipeline_tag: text-generation
|
|
---
|
|
|
|
# OTel-LLM-4B-IT
|
|
|
|
**OTel-LLM-4B-IT** is a telecom-specialized language model fine-tuned on telecommunications domain data. It is part of the [OTel Family of Models](https://huggingface.co/collections/farbodtavakkoli/otel-llm), an open-source initiative to build industry-standard AI models for the global telecommunications sector.
|
|
|
|
## Model Details
|
|
|
|
| Attribute | Value |
|
|
|-----------|-------|
|
|
| **Base Model** | [google/gemma-3-4b-it](https://huggingface.co/google/gemma-3-4b-it) |
|
|
| **Parameters** | 4B |
|
|
| **Training Method** | Full parameter fine-tuning |
|
|
| **Language** | English |
|
|
| **License** | Apache 2.0 |
|
|
|
|
## Training Data
|
|
|
|
The model was trained on high-quality telecom-focused data curated by 100+ domain experts from organizations including AT&T, Microsoft, AMD, GSMA, RelationalAI, Essential AI, Purdue University, Khalifa University, University of Leeds, Yale University, The University of Texas at Dallas, NetoAI, and MantisNLP.
|
|
|
|
**Data Sources:**
|
|
- GSMA Permanent Reference Documents
|
|
- 3GPP Specifications
|
|
- O-RAN Documentation
|
|
- RFC Series
|
|
- eSIM, terminals, security, networks, roaming, APIs
|
|
- Industry whitepapers and telecom academic papers
|
|
|
|
## Intended Use
|
|
|
|
The OTel model family is designed to power end-to-end Retrieval-Augmented Generation (RAG) pipelines for telecommunications. The three model types serve complementary roles:
|
|
|
|
1. **Embedding** — Retrieve relevant chunks from telecom specifications, standards, and documentation.
|
|
2. **Reranker** — Re-score and prioritize the retrieved chunks for relevance.
|
|
3. **LLM** — Generate accurate responses grounded in the retrieved context.
|
|
|
|
Users can deploy the full pipeline or use individual models independently based on their needs.
|
|
|
|
**Note:** The LLMs include abstention training — if the model does not receive sufficient context, it will decline to answer rather than hallucinate. This means the models are optimized for context-grounded generation, not open-ended question answering.
|
|
|
|
## Related Models
|
|
|
|
### Language Models
|
|
- [OTel LLM Collection](https://huggingface.co/collections/farbodtavakkoli/otel-llm)
|
|
|
|
### Embedding Models
|
|
- [OTel Embedding Collection](https://huggingface.co/collections/farbodtavakkoli/otel-embedding)
|
|
|
|
### Reranker Models
|
|
- [OTel Reranker Collection](https://huggingface.co/collections/farbodtavakkoli/otel-reranker)
|
|
|
|
## Related Datasets
|
|
|
|
- [OTel-Embedding](https://huggingface.co/datasets/farbodtavakkoli/OTel-Embedding)
|
|
- [OTel-Safety](https://huggingface.co/datasets/farbodtavakkoli/OTel-Safety)
|
|
- [OTel-LLM](https://huggingface.co/datasets/farbodtavakkoli/OTel-LLM)
|
|
- [OTel-Reranker](https://huggingface.co/datasets/farbodtavakkoli/OTel-Reranker)
|
|
|
|
## Training Infrastructure
|
|
|
|
- **Framework**: ScalarLM (GPU-agnostic)
|
|
- **Compute**: AMD and NVIDIA GPUs.
|
|
|
|
## Citation
|
|
|
|
```bibtex
|
|
@misc{otel2026,
|
|
title={OTel: Open Telco AI Models},
|
|
author={Tavakkoli, Farbod and Diamos, Gregory and Paulk, Roderic and Terrazas, Jorden},
|
|
year={2026},
|
|
url={https://huggingface.co/farbodtavakkoli}
|
|
}
|
|
```
|
|
|
|
## Contact
|
|
|
|
If you have any technical questions, please feel free to reach out to farbod.tavakkoli@att.com or farbodtavakoli@gmail.com
|