metadata
layout: default
title: CLaRa Documentation
CLaRa Documentation
Welcome to the CLaRa documentation! This site provides comprehensive guides and references for using CLaRa.
What is CLaRa?
CLaRa (Continuous Latent Reasoning) is a unified framework for retrieval-augmented generation that performs embedding-based compression and joint optimization in a shared continuous space.
Documentation
- Getting Started - Installation and quick start guide
- Training Guide - Detailed instructions for all three training stages including data formats
- Inference Guide - How to use CLaRa models for inference
Quick Links
- GitHub Repository: github.com/apple/ml-CLaRa
- Main README: ../README.md
- Model Checkpoints: Hugging Face (Coming Soon)
Overview
CLaRa uses a three-stage training approach:
- Stage 1: Compression Pretraining - Learn effective document compression
- Stage 2: Compression Instruction Tuning - Adapt for downstream QA tasks
- Stage 3: End-to-End Fine-tuning (CLaRa) - Joint retrieval and generation optimization
For more details, see the Training Guide.
Citation
If you use CLaRa in your research, please cite:
@article{clara2024,
title={CLaRa: Unified Retrieval-Augmented Generation with Compression},
author={[Authors]},
journal={[Journal]},
year={2024},
eprint={XXXX.XXXXX},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/XXXX.XXXXX}
}