Remove arXiv ID from metadata and add links to GitHub and Project Page
Browse filesThis PR improves the model card by:
1. Removing the `arxiv` field from the YAML metadata (as Hugging Face guidelines recommend keeping it in the markdown section instead).
2. Adding direct links to the official GitHub repository and project page at the top of the model card for better visibility.
README.md
CHANGED
|
@@ -1,5 +1,4 @@
|
|
| 1 |
---
|
| 2 |
-
arxiv: "2607.02375"
|
| 3 |
license: apache-2.0
|
| 4 |
pipeline_tag: text-to-image
|
| 5 |
tags:
|
|
@@ -20,6 +19,8 @@ over a curated teacher reference. One forward pass at 512² (≈0.15–0.3 s/ima
|
|
| 20 |
**This 1-step student matches or exceeds its 4-step teacher on all three eval axes** (standard-mmdet GenEval
|
| 21 |
composition + PickScore human-preference proxy).
|
| 22 |
|
|
|
|
|
|
|
| 23 |
## Results (checkpoint = step 180)
|
| 24 |
|
| 25 |
Standard-mmdet GenEval (553 prompts, avg over 6 tasks) + PickScore (raw `logit_scale·cos`), vs the teacher:
|
|
@@ -28,7 +29,7 @@ Standard-mmdet GenEval (553 prompts, avg over 6 tasks) + PickScore (raw `logit_s
|
|
| 28 |
|---|---|---|---|
|
| 29 |
| naive klein @ 1 step (no distillation) | ~0.42 | 19.95 | 20.11 |
|
| 30 |
| 4-step FLUX.2 klein teacher | 0.7944 | 22.576 | 21.848 |
|
| 31 |
-
| **this — 1-step RDM (s180)** | **0.8258** | **22.755** | 21.817 |
|
| 32 |
|
| 33 |
Distillation lifts the 1-step model from a broken floor (PickScore-cv 19.95) to **above the 4-step teacher**
|
| 34 |
on GenEval (+3.1 pp) and PickScore-COCO-val (+0.18), and to parity on Pick-a-Pic — i.e. it recovers the
|
|
@@ -101,4 +102,4 @@ Paper: [arXiv:2607.02375](https://arxiv.org/abs/2607.02375) · [Hugging Face Pap
|
|
| 101 |
```
|
| 102 |
|
| 103 |
## License
|
| 104 |
-
Derived from **FLUX.2 klein** (Apache-2.0); released under Apache-2.0.
|
|
|
|
| 1 |
---
|
|
|
|
| 2 |
license: apache-2.0
|
| 3 |
pipeline_tag: text-to-image
|
| 4 |
tags:
|
|
|
|
| 19 |
**This 1-step student matches or exceeds its 4-step teacher on all three eval axes** (standard-mmdet GenEval
|
| 20 |
composition + PickScore human-preference proxy).
|
| 21 |
|
| 22 |
+
[💻 Code](https://github.com/vita-epfl/RDM) · [📖 Paper](https://huggingface.co/papers/2607.02375) · [🌐 Project Page](https://alan-lanfeng.github.io/rdm/)
|
| 23 |
+
|
| 24 |
## Results (checkpoint = step 180)
|
| 25 |
|
| 26 |
Standard-mmdet GenEval (553 prompts, avg over 6 tasks) + PickScore (raw `logit_scale·cos`), vs the teacher:
|
|
|
|
| 29 |
|---|---|---|---|
|
| 30 |
| naive klein @ 1 step (no distillation) | ~0.42 | 19.95 | 20.11 |
|
| 31 |
| 4-step FLUX.2 klein teacher | 0.7944 | 22.576 | 21.848 |
|
| 32 |
+
| **this — 1-step RDM (s180)** | **0.8258** | **22.755** | **21.817** |
|
| 33 |
|
| 34 |
Distillation lifts the 1-step model from a broken floor (PickScore-cv 19.95) to **above the 4-step teacher**
|
| 35 |
on GenEval (+3.1 pp) and PickScore-COCO-val (+0.18), and to parity on Pick-a-Pic — i.e. it recovers the
|
|
|
|
| 102 |
```
|
| 103 |
|
| 104 |
## License
|
| 105 |
+
Derived from **FLUX.2 klein** (Apache-2.0); released under Apache-2.0.
|