Datasets:
Commit ·
b5d745e
1
Parent(s): a5f32cd
Update README.md
Browse files
README.md
CHANGED
|
@@ -34,16 +34,6 @@ The dataset was introduced as part of the paper: [Human Learning by Model Feedba
|
|
| 34 |
[ourpaper]: https://arxiv.org/abs/2311.12131 "markdown our paper"
|
| 35 |
|
| 36 |
|
| 37 |
-
|
| 38 |
-
## Dataset Details
|
| 39 |
-
|
| 40 |
-
### Dataset Description
|
| 41 |
-
|
| 42 |
-
<!-- Provide a longer summary of what this dataset is. -->
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
- **Language(s) (NLP):** English
|
| 46 |
-
|
| 47 |
### Dataset Sources
|
| 48 |
|
| 49 |
<!-- Provide the basic links for the dataset. -->
|
|
@@ -63,14 +53,14 @@ Main Columns:
|
|
| 63 |
- 'timestamp' - a timestamp of the prompt creation
|
| 64 |
- 'label' - Ture whether an image that was generated based on that prompt was upscaled, otherwise False.
|
| 65 |
- 'id' - unique id of the prompt
|
| 66 |
-
- 'url_png' -
|
| 67 |
- 'main_content' - prefix of the prompt, without trailing magic-words
|
| 68 |
- 'concreteness' - concreteness score, based on the [this paper][concpaper]
|
| 69 |
- 'word_len' - the number of words
|
| 70 |
- 'repeat_words' - the occurrences of each word that appears more than once in the prompt, excluding stop words.
|
| 71 |
- 'reapeat_words_ratio' - repeat_words / word_len
|
| 72 |
- 'perplexity' - the perplexity GPT-2 assigns to each prompt.
|
| 73 |
-
- 'caption_0-3' - captions that were generated by the BLIP-2 model
|
| 74 |
- 'phase' - train/test split, as was used to train image/text classifiers
|
| 75 |
- 'magic_ratio' - the percentage of words that were recognized as magic words in the prompt
|
| 76 |
- 'thread_id' - the id of the thread
|
|
|
|
| 34 |
[ourpaper]: https://arxiv.org/abs/2311.12131 "markdown our paper"
|
| 35 |
|
| 36 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 37 |
### Dataset Sources
|
| 38 |
|
| 39 |
<!-- Provide the basic links for the dataset. -->
|
|
|
|
| 53 |
- 'timestamp' - a timestamp of the prompt creation
|
| 54 |
- 'label' - Ture whether an image that was generated based on that prompt was upscaled, otherwise False.
|
| 55 |
- 'id' - unique id of the prompt
|
| 56 |
+
- 'url_png' - link to the generated images (a 4-grid version)
|
| 57 |
- 'main_content' - prefix of the prompt, without trailing magic-words
|
| 58 |
- 'concreteness' - concreteness score, based on the [this paper][concpaper]
|
| 59 |
- 'word_len' - the number of words
|
| 60 |
- 'repeat_words' - the occurrences of each word that appears more than once in the prompt, excluding stop words.
|
| 61 |
- 'reapeat_words_ratio' - repeat_words / word_len
|
| 62 |
- 'perplexity' - the perplexity GPT-2 assigns to each prompt.
|
| 63 |
+
- 'caption_0-3' - captions that were generated by the BLIP-2 model, with the 4 created images as its inputs.
|
| 64 |
- 'phase' - train/test split, as was used to train image/text classifiers
|
| 65 |
- 'magic_ratio' - the percentage of words that were recognized as magic words in the prompt
|
| 66 |
- 'thread_id' - the id of the thread
|