Update README.md
Browse files
README.md
CHANGED
|
@@ -4,6 +4,32 @@ task_categories:
|
|
| 4 |
- text-generation
|
| 5 |
language:
|
| 6 |
- en
|
|
|
|
| 7 |
size_categories:
|
| 8 |
- 1K<n<10K
|
| 9 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
- text-generation
|
| 5 |
language:
|
| 6 |
- en
|
| 7 |
+
- zh
|
| 8 |
size_categories:
|
| 9 |
- 1K<n<10K
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Dataset Card for LongbenchSeg
|
| 13 |
+
|
| 14 |
+
<!-- Provide a quick summary of the dataset. -->
|
| 15 |
+
|
| 16 |
+
The segmented version of [Longbench](https://huggingface.co/datasets/zai-org/LongBench) using the trained [segmenter](https://huggingface.co/Syon-Li/Qwen3-4B-Instruct-2507-Segmenter) from the paper Towards Generalization of Block Attention via Automatic Segmentation and Block Distillation.
|
| 17 |
+
|
| 18 |
+
We set the recursion depth to 1 and use the threshold value of 0.4.
|
| 19 |
+
|
| 20 |
+
## Dataset Details
|
| 21 |
+
|
| 22 |
+
### Dataset Description
|
| 23 |
+
|
| 24 |
+
<!-- Provide a longer summary of what this dataset is. -->
|
| 25 |
+
|
| 26 |
+
The newly introduced segmentation columns are:
|
| 27 |
+
- chunks: The segmented chunks.
|
| 28 |
+
- cut_prob: The corresponding segmenting probability for each candidate cut point.
|
| 29 |
+
- parallel_degree: The number of segmented chunks.
|
| 30 |
+
|
| 31 |
+
|
| 32 |
+
If you find this dataset useful, please cite:
|
| 33 |
+
```
|
| 34 |
+
|
| 35 |
+
```
|