---
license: mit
tags:
- reinforcement-learning
- reward-model
- robotics
- metaworld
- learning-from-videos
---

# TimeRewarder — MetaWorld checkpoints

Pretrained **TimeRewarder** checkpoints for 10 MetaWorld tasks, from the paper:

**[TimeRewarder: Learning Dense Reward from Passive Videos via Frame-wise Temporal Distance](https://arxiv.org/abs/2509.26627)**
Yuyang Liu\*, Chuan Wen\*, Yihang Hu, Dinesh Jayaraman, Yang Gao†

[🌐 Project Page](https://timerewarder.github.io/) · [📄 Paper](https://arxiv.org/abs/2509.26627) · [💻 Code](https://github.com/CowAndSheep/TimeRewader) · [🤗 Demos](https://huggingface.co/datasets/CowAndSheep/timerewarder-demos)

TimeRewarder learns a dense reward from passive (action-free) videos by predicting the frame-wise temporal distance between frames; the per-frame progress prediction is used directly as the reward for downstream RL.

## Contents

One vision-only checkpoint per task (`<task>_20bins.pth`, 20 discrete progress bins):
`basketball`, `button-press-topdown`, `disassemble`, `door-open`, `drawer-open`, `lever-pull`, `plate-slide`, `stick-push`, `window-close`, `window-open`.

```bash
huggingface-cli download CowAndSheep/timerewarder --local-dir models/ckpt
```

Drop the `.pth` files into `models/ckpt/` and run the downstream RL as described in the [code repository](https://github.com/CowAndSheep/TimeRewader).

## Citation

```bibtex
@article{liu2025timerewarder,
  title={TimeRewarder: Learning Dense Reward from Passive Videos via Frame-wise Temporal Distance},
  author={Liu, Yuyang and Wen, Chuan and Hu, Yihang and Jayaraman, Dinesh and Gao, Yang},
  journal={arXiv preprint arXiv:2509.26627},
  year={2025}
}
```