We will release the code and data later. Stay tuned!

If you are interested in testing your model with our benchmark or using our constructed training data in your long-context post-training, please contact miao.li@ed.ac.uk.