Add model card
#1
by nielsr HF Staff - opened
README.md
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
pipeline_tag: translation
|
| 4 |
+
---
|
| 5 |
+
|
| 6 |
+
# ELF: Embedded Language Flows
|
| 7 |
+
|
| 8 |
+
ELF is a class of continuous diffusion language models (DLMs) in continuous embedding space based on continuous-time Flow Matching. Unlike existing 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).
|
| 9 |
+
|
| 10 |
+
Experiments show that ELF substantially outperforms leading discrete and continuous DLMs, achieving better generation quality with fewer sampling steps.
|
| 11 |
+
|
| 12 |
+
- **Paper:** [ELF: Embedded Language Flows](https://huggingface.co/papers/2605.10938)
|
| 13 |
+
- **Repository:** [https://github.com/lillian039/ELF](https://github.com/lillian039/ELF)
|
| 14 |
+
|
| 15 |
+
## Citation
|
| 16 |
+
|
| 17 |
+
```bibtex
|
| 18 |
+
@article{elf2026,
|
| 19 |
+
title={ELF: Embedded Language Flows},
|
| 20 |
+
author={Hu, Keya and Qiu, Linlu and Lu, Yiyang and Zhao, Hanhong bit, Tianhong Li, Yoon Kim, Jacob Andreas, and Kaiming He},
|
| 21 |
+
journal={arXiv preprint arXiv:2605.10938},
|
| 22 |
+
year={2026}
|
| 23 |
+
}
|
| 24 |
+
```
|