diff --git a/L2D/README.md b/L2D/README.md
index 0d2ab3f..f42c511 100644
--- a/L2D/README.md
+++ b/L2D/README.md
@@ -9,7 +9,11 @@
 
 #### 1. Meta-learning implementation: 
 
-Please ref to supervised_maml.py
+Please ref to `supervised_maml.py`
+
+*Note: we do not implement the task scheduler since the data generation process in L2D is not flexible as POMO.*
+
+*Note: based on our exps, the training of the first-order approximation is not stable in L2D. Nevertheless, unlike POMO, the second-order one for L2D is computationally efficient.*
 
 #### 2. How to run? 
 
diff --git a/L2D/code/supervised_maml.py b/L2D/code/supervised_maml.py
index 65280de..7b77166 100644
--- a/L2D/code/supervised_maml.py
+++ b/L2D/code/supervised_maml.py
@@ -571,7 +571,7 @@ def log(text, **kwargs):
         if step % args.n_step_generate == 0 and (step > 0 or args.generate_step_zero):
             generate(args, d_generate, net, step)
         if step == args.n_steps: break
-        if step > 10000:  # simple curriculm learning strategy 
+        if step > 10000:  # simple curriculm learning strategy, note: could simply use random task scheduler instead,
             start, end = 75000, d.N
 
         opt.zero_grad()
diff --git a/README.md b/README.md
index b207469..e158513 100644
--- a/README.md
+++ b/README.md
@@ -1 +1,54 @@
-# TBA
\ No newline at end of file
+<h1 align="center"> Towards Omni-generalizable Neural Methods for Vehicle Routing Problems </h1>
+
+<p align="center">
+      <a href="https://openreview.net/forum?id=PlFBOnVOFg"><img src="https://img.shields.io/badge/OpenReview-PDF-brightgreen" alt="Paper"></a>&nbsp;&nbsp;&nbsp;&nbsp;<a href="https://icml.cc/Conferences/2023"><img alt="License" src="https://img.shields.io/static/v1?label=ICML'23&message=Hawaii&color=9cf"></a>&nbsp;&nbsp;&nbsp;&nbsp;<a href="https://hits.seeyoufarm.com"><img src="https://hits.seeyoufarm.com/api/count/incr/badge.svg?url=https%3A%2F%2Fgithub.com%2FRoyalSkye%2FOmni-VRP&count_bg=%233DC4C8&title_bg=%23555555&icon=github.svg&icon_color=%23E7E7E7&title=stats&edge_flat=false"/></a>&nbsp;&nbsp;&nbsp;&nbsp;<a href="https://github.com/RoyalSkye/Omni-VRP/blob/main/LICENSE"><img src="https://img.shields.io/badge/License-MIT-orange" alt="Paper"></a>
+  </p>
+
+The PyTorch Implementation of *ICML 2023 Poster -- "Towards Omni-generalizable Neural Methods for Vehicle Routing Problems"* by [Jianan Zhou](https://royalskye.github.io), [Yaoxin Wu](https://research.tue.nl/en/persons/yaoxin-wu), [Wen Song](https://songwenas12.github.io), [Zhiguang Cao](https://zhiguangcaosg.github.io), [Jie Zhang](https://personal.ntu.edu.sg/zhangj).
+
+<p align="center"><img src="./imgs/overview.png" width=95%></p>
+
+### TL;DR
+
+This paper studies a challenging yet realistic setting, which considers generalization across both size and distribution (a.k.a. omni-generalization) of neural methods in VRPs. Technically, a general meta-learning framework is developed to tackle it.
+
+### TODO
+
+- [ ] Finish Dependencies & How to Run.
+- [ ] Camera-ready.
+- [ ] Slide and Poster.
+- [ ] Release Review.
+
+### Dependencies
+
+
+
+### How to Run
+
+
+
+### Reviews
+
+We would like to thank the anonymous reviewers and (S)ACs of ICML 2023 for their constructive comments and recommendation. We will share the reviews later.
+
+### Acknowledgments
+
+Thank the following repositories, which are baselines of our code:
+
+* https://github.com/wouterkool/attention-learn-to-route
+* https://github.com/yd-kwon/POMO
+* https://github.com/mit-wu-lab/learning-to-delegate
+
+### Citation
+
+If you find our paper and code useful, please cite our paper:
+
+```tex
+@inproceedings{zhou2023towards,
+title       ={Towards Omni-generalizable Neural Methods for Vehicle Routing Problems},
+author      ={Jianan Zhou and Yaoxin Wu and Wen Song and Zhiguang Cao and Jie Zhang},
+booktitle   ={International Conference on Machine Learning},
+year        ={2023}
+}
+```
+
diff --git a/imgs/overview.png b/imgs/overview.png
new file mode 100644
index 0000000..6d6e342
Binary files /dev/null and b/imgs/overview.png differ