# Mirage

- [Mirage](https://mirage-eval.com/index.md): A modular framework for evaluating object hallucination in vision-language models.
- [Getting started](https://mirage-eval.com/getting-started.md): Set up Mirage and verify the demo path without downloading a model.
- [Datasets](https://mirage-eval.com/datasets.md): Prepare the local COCO and Visual Genome files used by Mirage.
- [Run on an HPC cluster](https://mirage-eval.com/hpc.md): Prepare Mirage, model weights, and datasets, then submit reproducible GPU evaluations with SLURM.
- [Running evaluations](https://mirage-eval.com/running-evaluations.md): Use demo.py to select a dataset, metric, model, and output file.
- [Metrics and models](https://mirage-eval.com/metrics-and-models.md): Implemented metric paths and model backends available in the demo.
- [Maintain the docs](https://mirage-eval.com/maintaining-docs.md): How Mirage documentation is organized, checked, previewed, and published.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.