---
title: AI Evaluation Toolkit
emoji: 🎯
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 4.44.0
app_file: app.py
pinned: true
short_description: RLHF rating, content policy scoring, obs/inference
---

# AI Evaluation Toolkit

Interactive demos of the AI training data quality control workflows from [github.com/LaelaZorana](https://github.com/LaelaZorana).

Three tools:
1. **RLHF Pairwise Rater** — Rate AI responses on 4 axes with self-consistency check
2. **Content Policy Rater** — Score text against a policy rubric with per-criterion reasoning
3. **Observation vs Inference** — Practice keeping observations clean of conclusions

Built by [Laela Zorana](https://github.com/LaelaZorana) | [HuggingFace](https://huggingface.co/LaelaZ) | [Kaggle](https://kaggle.com/laelazorana)