Add pipeline tag and paper/code links
#1
by nielsr HF Staff - opened
README.md
CHANGED
|
@@ -1,13 +1,14 @@
|
|
| 1 |
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
library_name: pytorch
|
| 4 |
base_model: facebook/pe-av-large
|
|
|
|
|
|
|
|
|
|
| 5 |
tags:
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
---
|
| 12 |
|
| 13 |
# SyncReward
|
|
@@ -16,7 +17,9 @@ SyncReward scores the **audio-visual synchronization** of a generated video with
|
|
| 16 |
reward in **[0, 2]** (0 = not synchronized, 2 = well synchronized) and is trained to match human sync
|
| 17 |
judgments on clips from text-to-audio-video generators.
|
| 18 |
|
| 19 |
-
Paper
|
|
|
|
|
|
|
| 20 |
|
| 21 |
## Models
|
| 22 |
|
|
@@ -56,7 +59,7 @@ hf download Guan123/SyncReward --include 'syncreward-stage2/*' --local-dir check
|
|
| 56 |
(cd checkpoints/syncreward-stage2 && sha256sum -c SHA256SUMS)
|
| 57 |
```
|
| 58 |
|
| 59 |
-
Loading uses the SyncReward code (repository link
|
| 60 |
|
| 61 |
```python
|
| 62 |
from syncreward.model import build_reward_model
|
|
@@ -106,4 +109,14 @@ Apache-2.0 (see [LICENSE](LICENSE)). The PE-AV backbone `facebook/pe-av-large` i
|
|
| 106 |
|
| 107 |
## Citation
|
| 108 |
|
| 109 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
|
|
|
|
|
|
| 2 |
base_model: facebook/pe-av-large
|
| 3 |
+
library_name: pytorch
|
| 4 |
+
license: apache-2.0
|
| 5 |
+
pipeline_tag: other
|
| 6 |
tags:
|
| 7 |
+
- audio-video
|
| 8 |
+
- audio-visual
|
| 9 |
+
- synchronization
|
| 10 |
+
- reward-model
|
| 11 |
+
- video-generation
|
| 12 |
---
|
| 13 |
|
| 14 |
# SyncReward
|
|
|
|
| 17 |
reward in **[0, 2]** (0 = not synchronized, 2 = well synchronized) and is trained to match human sync
|
| 18 |
judgments on clips from text-to-audio-video generators.
|
| 19 |
|
| 20 |
+
Paper: [SyncReward: Human-Aligned Synchronization Reward for Audio-Visual Generation](https://huggingface.co/papers/2610.16338)
|
| 21 |
+
|
| 22 |
+
Code: [https://github.com/WeChatCV/SyncReward](https://github.com/WeChatCV/SyncReward)
|
| 23 |
|
| 24 |
## Models
|
| 25 |
|
|
|
|
| 59 |
(cd checkpoints/syncreward-stage2 && sha256sum -c SHA256SUMS)
|
| 60 |
```
|
| 61 |
|
| 62 |
+
Loading uses the SyncReward code (repository link above):
|
| 63 |
|
| 64 |
```python
|
| 65 |
from syncreward.model import build_reward_model
|
|
|
|
| 109 |
|
| 110 |
## Citation
|
| 111 |
|
| 112 |
+
```bibtex
|
| 113 |
+
@misc{guan2026syncrewardhumanalignedsynchronizationreward,
|
| 114 |
+
title={SyncReward: Human-Aligned Synchronization Reward for Audio-Visual Generation},
|
| 115 |
+
author={Kaisi Guan and Ying Ba and Wenjing Wang and Xihua Wang and Chen Li and Jing Lyu and Ruihau Song},
|
| 116 |
+
year={2026},
|
| 117 |
+
eprint={2610.16338},
|
| 118 |
+
archivePrefix={arXiv},
|
| 119 |
+
primaryClass={cs.CV},
|
| 120 |
+
url={https://arxiv.org/abs/2610.16338},
|
| 121 |
+
}
|
| 122 |
+
```
|