Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -59,4 +59,20 @@ configs:
|
|
| 59 |
path: dom-element/validation-*
|
| 60 |
- split: test
|
| 61 |
path: dom-element/test-*
|
|
|
|
|
|
|
|
|
|
| 62 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 59 |
path: dom-element/validation-*
|
| 60 |
- split: test
|
| 61 |
path: dom-element/test-*
|
| 62 |
+
license: cc-by-4.0
|
| 63 |
+
source_datasets:
|
| 64 |
+
- osunlp/Mind2Web
|
| 65 |
---
|
| 66 |
+
|
| 67 |
+
# mind2web-dom
|
| 68 |
+
|
| 69 |
+
Mind2Web public training trajectories as compact HTML-only actions and native MCQs.
|
| 70 |
+
|
| 71 |
+
Dataset annotations are CC BY 4.0: https://huggingface.co/datasets/osunlp/Mind2Web
|
| 72 |
+
No official benchmark test files are downloaded or republished. Derived 80/10/10
|
| 73 |
+
train/validation/test partitions group whole trajectories and identical cleaned DOM
|
| 74 |
+
snapshots. They are internal holdouts, not the official Mind2Web benchmark splits.
|
| 75 |
+
Only past actions enter context. Four-way MC uses the original target and three
|
| 76 |
+
deterministically sampled native negatives, excluding all alternative positives.
|
| 77 |
+
|
| 78 |
+
Original data: [osunlp/Mind2Web](https://huggingface.co/datasets/osunlp/Mind2Web). Repackaged as parquet for [tasksource](https://github.com/sileod/tasksource) by [scripts/repackage_dataset/](https://github.com/sileod/tasksource/tree/main/scripts/repackage_dataset).
|