Datasets:
Add contributor and clarify manual construction and domain statistics
Browse files
README.md
CHANGED
|
@@ -106,21 +106,6 @@ are explicitly retained so evaluation is reproducible on this collection.
|
|
| 106 |
|
| 107 |
## Dataset statistics
|
| 108 |
|
| 109 |
-
| Statistic | Full release |
|
| 110 |
-
| --- | --- |
|
| 111 |
-
| Human-recorded tasks | 549 |
|
| 112 |
-
| Websites | 144 |
|
| 113 |
-
| Recorded steps in reference-task metadata | 14,042 |
|
| 114 |
-
| Distributed scored steps / screenshots | 12,905 |
|
| 115 |
-
| Mean / median scored steps per task | 23.51 / 21 |
|
| 116 |
-
| Scored steps per task, minimum–maximum | 3–100 |
|
| 117 |
-
| Steps with primary target boxes | 7,197 |
|
| 118 |
-
| Steps with a primary or alternative target box | 7,580 |
|
| 119 |
-
| Steps with alternative valid actions | 755 |
|
| 120 |
-
| Alternative valid actions | 756 |
|
| 121 |
-
| Original / concise instruction records | 549 / 549 |
|
| 122 |
-
| Viewer examples | 10 |
|
| 123 |
-
|
| 124 |
### Recorded operation distribution
|
| 125 |
|
| 126 |
| Category | Count | Share |
|
|
@@ -138,29 +123,6 @@ step file contains the 12,905 scored steps; `GOTO`, `WAIT`, and `FINISH` are
|
|
| 138 |
excluded from step scoring. `statistics.json` includes per-website counts,
|
| 139 |
task categories, trajectory-length statistics, and alternative-action counts.
|
| 140 |
|
| 141 |
-
## Files and fields
|
| 142 |
-
|
| 143 |
-
| File | Contents |
|
| 144 |
-
| --- | --- |
|
| 145 |
-
| `multimodal_mind2web_ko.tar.gz` | Scored steps, task metadata, instruction variants, and screenshots |
|
| 146 |
-
| `examples/test.parquet` | Ten viewer examples with embedded images |
|
| 147 |
-
| `examples/samples.jsonl` | Complete example records |
|
| 148 |
-
| `examples/images/` | The corresponding example PNGs |
|
| 149 |
-
| `statistics.json` | Full-release aggregate statistics |
|
| 150 |
-
| `evaluate.py` | Reference element/step/task scorer |
|
| 151 |
-
|
| 152 |
-
Extraction creates `multimodal_mind2web_ko/steps.jsonl`, `tasks.jsonl`,
|
| 153 |
-
`tasks_rewrite.jsonl`, and `images/`. The step records include `sample_id`,
|
| 154 |
-
`task_id`, `instruction`, `image`, `image_size`, `action`, `previous_actions`,
|
| 155 |
-
`scored`, target geometry, and `alt_actions`. Image paths are relative to that
|
| 156 |
-
extracted directory. The preview exposes the structured action as `action_json`
|
| 157 |
-
and places `task_id` in the final column.
|
| 158 |
-
|
| 159 |
-
Pixel coordinates, [0, 1] normalized fields, and the original `*_999` grid
|
| 160 |
-
fields are preserved. The scorer evaluates in the 0–999 grid, converting
|
| 161 |
-
pixel or [0, 1] predictions to that coordinate system. Use the same convention
|
| 162 |
-
when comparing scores.
|
| 163 |
-
|
| 164 |
## Download and preview
|
| 165 |
|
| 166 |
This repository is private. Authenticate with a Hugging Face account that has
|
|
|
|
| 106 |
|
| 107 |
## Dataset statistics
|
| 108 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 109 |
### Recorded operation distribution
|
| 110 |
|
| 111 |
| Category | Count | Share |
|
|
|
|
| 123 |
excluded from step scoring. `statistics.json` includes per-website counts,
|
| 124 |
task categories, trajectory-length statistics, and alternative-action counts.
|
| 125 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 126 |
## Download and preview
|
| 127 |
|
| 128 |
This repository is private. Authenticate with a Hugging Face account that has
|