README.md

---
license: mit
pipeline_tag: text-generation
---

# ELF: Embedded Language Flows

ELF is a class of continuous diffusion language models based on continuous-time Flow Matching, presented in the paper [ELF: Embedded Language Flows](https://huggingface.co/papers/2605.10938).

Unlike existing diffusion language models (DLMs), ELF predominantly stays within the continuous embedding space until the final time step, where it maps to discrete tokens using a shared-weight network. This formulation makes it straightforward to adapt established techniques from image-domain diffusion models, such as classifier-free guidance (CFG).

- **Paper:** [ELF: Embedded Language Flows](https://huggingface.co/papers/2605.10938)
- **Repository:** [https://github.com/lillian039/ELF](https://github.com/lillian039/ELF)

## Inference and Evaluation

This checkpoint can be used with the official JAX implementation. To verify the setup or run evaluation, follow these steps:

### Initialization

```bash
git clone https://github.com/lillian039/ELF
cd ELF
pip install -r requirements.txt
```

### Running Evaluation

To evaluate the model on unconditional generation (OpenWebText):

```bash
cd src/

# For ELF-B (105M)
python eval.py \
    --config configs/training_configs/train_owt_ELF-B.yml \
    --checkpoint_path embedded-language-flows/ELF-B-owt
```

To evaluate on conditional tasks like summarization (XSum):

```bash
cd src/

python eval.py \
    --config configs/training_configs/train_xsum_ELF-B.yml \
    --checkpoint_path embedded-language-flows/ELF-B-xsum
```

## Citation

```bibtex
@article{elf2026,
  title={ELF: Embedded Language Flows},
  author={Hu, Keya and Qiu, Linlu and Lu, Yiyang and Zhao, Hanhong and Li, Tianhong and Kim, Yoon and Andreas, Jacob and He, Kaiming},
  journal={arXiv preprint arXiv:2605.10938},
  year={2026}
}
```