metadata

title: H4rmonyEval
emoji: 🌍
colorFrom: purple
colorTo: gray
sdk: gradio
sdk_version: 5.27.0
app_file: app.py
pinned: false
license: mit
short_description: Model Judge For Ecolinguistic Evaluation

🐦‍⬛ H4rmonyEval: Model Judge for Ecolinguistic Evaluation

H4rmonyEval is a specialised evaluation tool designed to assess language model outputs through an ecolinguistic perspective. It enables both single prompt evaluation and batch assessments via CSV, focusing on how well model responses align with ecological and sustainable narratives.

Hosted on Hugging Face Spaces, powered by Gradio.

🌿 Features

Single Q&A Evaluation
Quickly score individual prompt/answer pairs on a scale of 0–5.
Batch CSV Evaluation
Upload a CSV with questions (and optional answers) for automated scoring. Download detailed results.
Ecolinguistic-Aware Judging
Leveraging a Cohere language model with custom system instructions based on ecolinguistic principles.

📄 CSV Format

Required: question column.
Optional: answer column (if empty, the system generates answers using a selected model).

Example:

question	answer
What is green AI?	AI focused on...
Define ecological ethics

🎯 Scoring System

5 – Perfect, accurate, and ecologically aligned.
3 – Partially correct, limited ecological framing.
0 – Incorrect or ecologically harmful/misaligned.

Final scores are presented as percentages for batch evaluations.

⚙️ How It Works

Judge: Cohere command-r-plus model guided by custom instructions.
Answer Generation: Hugging Face Inference API (if no answer is provided).
Frontend: Gradio 5.27.0.

🚀 Quick Start

Upload your CSV or enter a single Q&A.
Click the appropriate evaluation button.
View your score and download results.

🔐 Environment Variables

COHERE_API_KEY – Required.
HF_TOKEN – (Optional) For Hugging Face Inference.

🌍 About H4rmony

Part of the H4rmony Project, promoting AI aligned with ecological justice and sustainability through language.

Learn more:
🔗 https://TheH4rmonyProject.org

🤝 Acknowledgements

Cohere for LLM infrastructure.
Hugging Face for Spaces and APIs.
Inspired by the work of the The H4rmony Project, University of Gloucestershire & International Ecolinguistics Association.

📜 License

MIT License.

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference