Update README.md

README.md

@@ -1,21 +1,26 @@
----
-license: mit
----
-**Scaling Robot Learning with Heterogeneous Pre-training**
+# 🦾 Heterogenous Pre-trained Transformers
+ 
+[Lirui Wang](https://liruiw.github.io/), [Xinlei Chen](https://xinleic.xyz/), [Jialiang Zhao](https://alanz.info/), [Kaiming He](https://people.csail.mit.edu/kaiming/)
 
-**Abstract**
+Neural Information Processing Systems (Spotlight), 2024
 
-One of the key roadblocks for training generalist robotic models today is heterogeneity. Previous robot learning methods often collect data to train with one specific embodiment for one task, which is expensive and prone to overfitting. This work studies the problem of learning policy representations through heterogeneous pre-training on robot data across different embodiments and tasks at scale. We propose Heterogeneous Pre-trained Transformers (HPT), which pre-train a large, shareable trunk of a policy neural network to learn a task and embodiment agnostic shared representation. This general architecture aligns the specific proprioception and vision inputs from distinct embodiments to a short sequence of tokens and then processes such tokens to map to control robots for different tasks.  Leveraging the recent large-scale multi-embodiment real-world robotic datasets as well as simulation and human video datasets, we show that pre-training robotic policies across heterogeneity can exhibit compelling scaling behaviors, to the extent of 52 distinct datasets and 1 billion parameter models. Pre-trained HPTs outperform previous methods and enhance the finetuned policy performance on unseen downstream tasks and environments in simulator benchmarks and real-world settings. 
+You can find more details on our [project page](https://liruiw.github.io/hpt). An alternative clean implementation of HPT in Hugging Face can also be found [here](https://github.com/liruiw/lerobot/tree/hpt_squash/lerobot/common/policies/hpt).
 
-**Citation**
 
+**TL;DR:** HPT aligns different embodiment to a shared latent space and investigates the scaling behaviors in policy learning. Put a scalable transformer in the middle of your policy and don’t train from scratch!
+
+
+If you find HPT useful in your research, please consider citing:
+```
 @inproceedings{wang2024hpt,
-author    = {Lirui Wang, Xinlei Chen, Jialiang Zhao, Russ Tedrake, Kaiming He},
-title     = {Scaling Robot Learning with Heterogeneous Pre-training},
-booktitle = {Arxiv},
+author    = {Lirui Wang, Xinlei Chen, Jialiang Zhao, Kaiming He},
+title     = {Scaling Proprioceptive-Visual Learning with Heterogeneous Pre-trained Transformers},
+booktitle = {Neurips},
 year      = {2024}
 }
+```
+
 
-**Contact**
+## Contact
 
-Lirui Wang (liruiw@mit.edu)
+If you have any questions, feel free to contact me through email (liruiw@mit.edu). Enjoy!