---
license: mit
language:
- en
base_model:
- ibm-granite/granite-3.0-1b-a400m-base
pipeline_tag: text-generation
library_name: transformers
---

![image/png](/static-proxy?url=https%3A%2F%2Fcdn-uploads.huggingface.co%2Fproduction%2Fuploads%2F66f5800df74ecff7c0b9b64f%2F_TWMClyq8hFuM1gryXaii.png)

We are developing a search engine that introduces a novel AI-driven truth and reliability scoring system, assigning each search result a truth parameter on a scale of 0 (lie) to 1 (absolute truth).

TrueGL_Granite is an LLM fine-tuned on the large set of articles on various topics. Advanced algorithms were used to generate negative samples (completely unreliable data).

Our GitHub repository (with the fine-tuning and inference code) is publicly available at https://github.com/AlgazinovAleksandr/TrueGL

Note that the project is created for educational and research purposes only and is not intended for commercial use. The data used for training and fine-tuning the AI models is either collected from open-sources or AI-generated and is not collected or used in any way that violates privacy or ethical guidelines.


If you find this useful consider citing:

@misc{ch2025truegl,
    title={TrueGL: A Truthful, Reliable, and Unified Engine for Grounded Learning in Full-Stack Search},
    author={Joydeep Chandra and Aleksandr Algazinov and Satyam Kumar Navneet and Rim El Filali and Matt Laing and Andrew Hanna},
    year={2025},
    eprint={2506.12072},
    archivePrefix={arXiv},
    primaryClass={cs.IR}
}