neelalex commited on
Commit
14aa1a8
1 Parent(s): 6e059a4

Adding train/test count numbers. TODO: descriptive details about the subsets?

Browse files
Files changed (1) hide show
  1. README.md +28 -11
README.md CHANGED
@@ -65,7 +65,7 @@ The Real-world Annotation for Few-shot Tasks (RAFT) dataset is an aggregation of
65
 
66
  ### Supported Tasks and Leaderboards
67
 
68
- - `text-classification`: Each subtask in RAFT is a text classification task, and the provided train and test sets can be used to submit to the (RAFT Leaderboard)[https://huggingface.co/spaces/ought/raft-leaderboard] To prevent overfitting and tuning on a held-out test set, the leaderboard is only evaluated once per week. Each task has its macro-f1 score calculated, then those scores are averaged to produce the overall leaderboard score.
69
 
70
  ### Languages
71
 
@@ -80,15 +80,15 @@ The only language intentionally included in the dataset is American English (en-
80
  | ----------- | ----------- |
81
  | Ade Corpus V2 | <pre>Sentence: No regional side effects were noted.<br>ID: 0<br>Label: 2</pre> |
82
  | Banking 77 | <pre>Query: Is it possible for me to change my PIN number?<br>ID: 0<br>Label: 23<br></pre> |
83
- | Terms Of Service | <pre>Sentence: Crowdtangle may change these terms of service, as described above, notwithstanding any provision to the contrary in any agreemen...<br>ID: 0<br>Label: 2<br></pre> |
84
- | Tai Safety Research | <pre>Title: Malign generalization without internal search<br>Abstract Note: In my last post, I challenged the idea that inner alignment failures should be explained by appealing to agents which perform ex...<br>Url: https://www.alignmentforum.org/posts/ynt9TD6PrYw6iT49m/malign-generalization-without-internal-search...<br>Publication Year: 2020<br>Item Type: blogPost<br>Author: Barnett, Matthew<br>Publication Title: AI Alignment Forum<br>ID: 0<br>Label: 1</pre> |
85
  | Neurips Impact Statement Risks | <pre>Paper title: Auto-Panoptic: Cooperative Multi-Component Architecture Search for Panoptic Segmentation...<br>Paper link: https://proceedings.neurips.cc/paper/2020/file/ec1f764517b7ffb52057af6df18142b7-Paper.pdf...<br>Impact statement: This work makes the first attempt to search for all key components of panoptic pipeline and manages to accomplish this via the p...<br>ID: 0<br>Label: 1</pre> |
 
86
  | Overruling | <pre>Sentence: in light of both our holding today and previous rulings in johnson, dueser, and gronroos, we now explicitly overrule dupree....<br>ID: 0<br>Label: 2<br></pre> |
 
87
  | Systematic Review Inclusion | <pre>Title: Prototyping and transforming facial textures for perception research...<br>Abstract: Wavelet based methods for prototyping facial textures for artificially transforming the age of facial images were described. Pro...<br>Authors: Tiddeman, B.; Burt, M.; Perrett, D.<br>Journal: IEEE Comput Graphics Appl<br>ID: 0<br>Label: 2</pre> |
88
- | One Stop English | <pre>Article: For 85 years, it was just a grey blob on classroom maps of the solar system. But, on 15 July, Pluto was seen in high resolution ...<br>ID: 0<br>Label: 3<br></pre> |
 
89
  | Tweet Eval Hate | <pre>Tweet: New to Twitter-- any men on here know what the process is to get #verified?...<br>ID: 0<br>Label: 2<br></pre> |
90
  | Twitter Complaints | <pre>Tweet text: @HMRCcustomers No this is my first job<br>ID: 0<br>Label: 2</pre> |
91
- | Semiconductor Org Types | <pre>Paper title: 3Gb/s AC-coupled chip-to-chip communication using a low-swing pulse receiver...<br>Organization name: North Carolina State Univ.,Raleigh,NC,USA<br>ID: 0<br>Label: 3<br></pre> |
92
 
93
 
94
  ### Data Fields
@@ -103,12 +103,29 @@ dataset = datasets.load_dataset("ought/raft", "ade_corpus_v2")
103
  label_info = dataset.features["Label"]
104
  # Use the int2str method to access the textual labels.
105
  print([label_info.int2str(i) for i in (0, 1, 2)])
106
- # >>> ['Unlabeled', 'ADE-related', 'not ADE-related']
107
  ```
108
 
109
  ### Data Splits
110
 
111
- [More Information Needed]
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
112
 
113
  ## Dataset Creation
114
 
@@ -168,15 +185,15 @@ RAFT aggregates many other datasets, each of which is provided under its own lic
168
  | ----------- | ----------- |
169
  | Ade Corpus V2 | Test |
170
  | Banking 77 | |
171
- | Terms Of Service | |
172
- | Tai Safety Research | |
173
  | Neurips Impact Statement Risks | |
 
174
  | Overruling | |
 
175
  | Systematic Review Inclusion | |
176
- | One Stop English | |
 
177
  | Tweet Eval Hate | |
178
  | Twitter Complaints | |
179
- | Semiconductor Org Types | |
180
 
181
 
182
 
 
65
 
66
  ### Supported Tasks and Leaderboards
67
 
68
+ - `text-classification`: Each subtask in RAFT is a text classification task, and the provided train and test sets can be used to submit to the [RAFT Leaderboard](https://huggingface.co/spaces/ought/raft-leaderboard) To prevent overfitting and tuning on a held-out test set, the leaderboard is only evaluated once per week. Each task has its macro-f1 score calculated, then those scores are averaged to produce the overall leaderboard score.
69
 
70
  ### Languages
71
 
 
80
  | ----------- | ----------- |
81
  | Ade Corpus V2 | <pre>Sentence: No regional side effects were noted.<br>ID: 0<br>Label: 2</pre> |
82
  | Banking 77 | <pre>Query: Is it possible for me to change my PIN number?<br>ID: 0<br>Label: 23<br></pre> |
 
 
83
  | Neurips Impact Statement Risks | <pre>Paper title: Auto-Panoptic: Cooperative Multi-Component Architecture Search for Panoptic Segmentation...<br>Paper link: https://proceedings.neurips.cc/paper/2020/file/ec1f764517b7ffb52057af6df18142b7-Paper.pdf...<br>Impact statement: This work makes the first attempt to search for all key components of panoptic pipeline and manages to accomplish this via the p...<br>ID: 0<br>Label: 1</pre> |
84
+ | One Stop English | <pre>Article: For 85 years, it was just a grey blob on classroom maps of the solar system. But, on 15 July, Pluto was seen in high resolution ...<br>ID: 0<br>Label: 3<br></pre> |
85
  | Overruling | <pre>Sentence: in light of both our holding today and previous rulings in johnson, dueser, and gronroos, we now explicitly overrule dupree....<br>ID: 0<br>Label: 2<br></pre> |
86
+ | Semiconductor Org Types | <pre>Paper title: 3Gb/s AC-coupled chip-to-chip communication using a low-swing pulse receiver...<br>Organization name: North Carolina State Univ.,Raleigh,NC,USA<br>ID: 0<br>Label: 3<br></pre> |
87
  | Systematic Review Inclusion | <pre>Title: Prototyping and transforming facial textures for perception research...<br>Abstract: Wavelet based methods for prototyping facial textures for artificially transforming the age of facial images were described. Pro...<br>Authors: Tiddeman, B.; Burt, M.; Perrett, D.<br>Journal: IEEE Comput Graphics Appl<br>ID: 0<br>Label: 2</pre> |
88
+ | Tai Safety Research | <pre>Title: Malign generalization without internal search<br>Abstract Note: In my last post, I challenged the idea that inner alignment failures should be explained by appealing to agents which perform ex...<br>Url: https://www.alignmentforum.org/posts/ynt9TD6PrYw6iT49m/malign-generalization-without-internal-search...<br>Publication Year: 2020<br>Item Type: blogPost<br>Author: Barnett, Matthew<br>Publication Title: AI Alignment Forum<br>ID: 0<br>Label: 1</pre> |
89
+ | Terms Of Service | <pre>Sentence: Crowdtangle may change these terms of service, as described above, notwithstanding any provision to the contrary in any agreemen...<br>ID: 0<br>Label: 2<br></pre> |
90
  | Tweet Eval Hate | <pre>Tweet: New to Twitter-- any men on here know what the process is to get #verified?...<br>ID: 0<br>Label: 2<br></pre> |
91
  | Twitter Complaints | <pre>Tweet text: @HMRCcustomers No this is my first job<br>ID: 0<br>Label: 2</pre> |
 
92
 
93
 
94
  ### Data Fields
 
103
  label_info = dataset.features["Label"]
104
  # Use the int2str method to access the textual labels.
105
  print([label_info.int2str(i) for i in (0, 1, 2)])
106
+ # ['Unlabeled', 'ADE-related', 'not ADE-related']
107
  ```
108
 
109
  ### Data Splits
110
 
111
+ There are two splits provided: train data and unlabeled test data.
112
+
113
+ The training examples were chosen at random. No attempt was made to ensure that classes were balanced or proportional in the training data -- indeed, the Banking 77 task with 77 different classes if used cannot fit all of its classes into the 50 training examples.
114
+
115
+ | Dataset | Train Size | Test Size | |
116
+ |--------------------------------|------------|-----------|---|
117
+ | Ade Corpus V2 | 50 | 5000 | |
118
+ | Banking 77 | 50 | 5000 | |
119
+ | Neurips Impact Statement Risks | 50 | 150 | |
120
+ | One Stop English | 50 | 518 | |
121
+ | Overruling | 50 | 2350 | |
122
+ | Semiconductor Org Types | 50 | 449 | |
123
+ | Systematic Review Inclusion | 50 | 2244 | |
124
+ | TAI Safety Research | 50 | 1639 | |
125
+ | Terms Of Service | 50 | 5000 | |
126
+ | Tweet Eval Hate | 50 | 2966 | |
127
+ | Twitter Complaints | 50 | 3399 | |
128
+ | **Total** | **550** | **28715** | |
129
 
130
  ## Dataset Creation
131
 
 
185
  | ----------- | ----------- |
186
  | Ade Corpus V2 | Test |
187
  | Banking 77 | |
 
 
188
  | Neurips Impact Statement Risks | |
189
+ | One Stop English | |
190
  | Overruling | |
191
+ | Semiconductor Org Types | |
192
  | Systematic Review Inclusion | |
193
+ | Tai Safety Research | |
194
+ | Terms Of Service | |
195
  | Tweet Eval Hate | |
196
  | Twitter Complaints | |
 
197
 
198
 
199