Datasets:

Modalities:
Image
Text
Formats:
parquet
Languages:
English
ArXiv:
Libraries:
Datasets
Dask
License:
boyuanzheng010 commited on
Commit
f3f3b45
1 Parent(s): 77e56b5

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +3 -4
README.md CHANGED
@@ -62,10 +62,9 @@ Multimodal-Mind2Web is the multimodal version of [Mind2Web](https://osu-nlp-grou
62
 
63
  ### Data Splits
64
  - train: 7775 actions from 1009 tasks.
65
- - test:
66
- - Cross Task: 1339 actions from 177 tasks. Tasks from the same website are seen during training.
67
- - Cross Website: 1019 actions from 142 tasks. Websites are not seen during training.
68
- - Cross Domain: 4060 actions from 694 tasks. Entire domains are not seen during training.
69
 
70
  The three test set splits have went through human verification to confirm element visibility and correct rendering for action prediction.
71
 
 
62
 
63
  ### Data Splits
64
  - train: 7775 actions from 1009 tasks.
65
+ - test_task: 1339 actions from 177 tasks. Tasks from the same website are seen during training.
66
+ - test_website: 1019 actions from 142 tasks. Websites are not seen during training.
67
+ - test_domain: 4060 actions from 694 tasks. Entire domains are not seen during training.
 
68
 
69
  The three test set splits have went through human verification to confirm element visibility and correct rendering for action prediction.
70