LongbenchSeg

# Dataset Card for LongbenchSeg <!-- Provide a quick summary of the dataset. --> The segmented version of [Longbench](https://huggingface.co/datasets/zai-org/LongBench) using the trained [segmenter](https://modelscope.cn/models/SyonLi/Qwen3-4B-Instruct-2507-Segmenter) from the paper [Towards Generalization of Block Attention via Automatic Segmentation and Block Distillation](https://arxiv.org/abs/2605.15913). We set the recursion depth to 1 and use the threshold value of 0.4. ## Dataset Details ### Dataset Description <!-- Provide a longer summary of what this dataset is. --> The newly introduced segmentation columns are: - chunks: The segmented chunks. - cut_prob: The corresponding segmenting probability for each candidate cut point. - parallel_degree: The number of segmented chunks. If you find this dataset useful, please cite: ``` @article{li2026towards, title={Towards Generalization of Block Attention via Automatic Segmentation and Block Distillation}, author={Li, Shuaiyi and Zhang, Zhisong and Wang, Yan and Zhu, Lei and Ma, Dongyang and Deng, Chenlong and Deng, Yang and Lam, Wai}, journal={arXiv preprint arXiv:2605.15913}, year={2026} } ```