Datasets:
ibm
/

Modalities:
Text
Formats:
csv
Languages:
English
Size:
< 1K
ArXiv:
Libraries:
Datasets
pandas
License:
javiccano commited on
Commit
48aab24
1 Parent(s): 8b8d3c2

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +25 -26
README.md CHANGED
@@ -83,33 +83,31 @@ The description of each key (when the instance contains two questions) is as fol
83
  - **tag:** Type of tag of the article (Inconsistent/Self-contradictory/Contradict-other).
84
  - **tagDate:** Date of the tag.
85
  - **tagReason:** Reason for the tag.
86
- - **annotationResult:** Results of the human data annotation.
87
- - **wikitag_label_valid:** Valid or invalid tag (Valid/Invalid).
88
- - **valid_comment:** Comment on the tag.
89
- - **paragraphA_article:** Title of article containing passage 1.
90
- - **paragraphA_information:** Relevant information of passage 1.
91
- - **paragraphA_information_standalone:** Decontextualized relevant information of passage 1.
92
- - **paragraphB_article:** Relevant information of passage 2.
93
- - **paragraphB_information_standalone:** Decontextualized relevant information of passage 2.
94
- - **wikitag_label_samepassage:** Boolean value stating whether passage 1 and passage 2 are the same (Same/Different).
95
- - **relevantInfo_comment_A:** Comment on the information of passage 1.
96
- - **relevantInfo_comment_B:** Comment on the information of passage 2.
97
- - **Contradict type I:** Contradiction type I focuses on the fine-grained semantics of the contradiction, e.g., date/time, location, language, etc.
98
- - **Contradict type II:** Contradiction type II focuses on the modality the contradiction. It describes the modality of passage 1 and passage 2, whether the information is from a piece of text, or from a row an infobox or a table.
99
- - **Contradict type III:** Contradiction type III focuses on the source the contradiction. It describes whether passage 1 and passage 2 are from the same article or not.
100
- - **Contradict type IV:** Contradiction type IV focuses on the reasoning aspect. It describes whether the contraction is explicit or implicit (Explicit/Implicit). Implicit contradiction requires some reasoning to understand why passage 1 and passage 2 are contradicted.
101
- - **taxonomy:** Array of key-values corresponding to contradict type I, contradict type II, contradict type III, and contradict type IV.
102
- - **question1:** Question 1 inferred from the contradiction.
103
- - **question1_answer1:** Gold answer to question 1 according to passage 1.
104
- - **question1_answer2:** Gold answer to question 1 according to passage 2.
105
- - **question2:** Question 2 inferred from the contradiction.
106
- - **question2_answer1:** Gold answer to question 2 according to passage 1.
107
- - **question2_answer2:** Gold answer to question 2 according to passage 2.
108
 
109
 
110
  ## Usage of the Dataset
111
 
112
- We provide the following starter code:
113
 
114
 
115
  ```python
@@ -148,15 +146,16 @@ parameters = TextGenerationParameters(
148
  max_new_tokens=250,
149
  min_new_tokens=1,
150
  decoding_method=DecodingMethod.GREEDY,
151
- # length_penalty=LengthPenalty(start_index=5, decay_factor=1.5),
152
  return_options=TextGenerationReturnOptions(
153
  # if ordered is False, you can use return_options to retrieve the corresponding prompt
154
  input_text=True,
155
  ),
156
  )
157
 
 
 
158
  testingUnits = load_testingdata()
159
- print("question size", len(testingUnits))
160
  generateAnswers_bam_models(testingUnits)
161
  ```
162
 
@@ -246,7 +245,7 @@ If this dataset is utilized in your research, kindly cite the following paper:
246
 
247
  **APA:**
248
 
249
- [More Information Needed]
250
 
251
  <!-- ## Glossary [optional] -->
252
 
 
83
  - **tag:** Type of tag of the article (Inconsistent/Self-contradictory/Contradict-other).
84
  - **tagDate:** Date of the tag.
85
  - **tagReason:** Reason for the tag.
86
+ - **wikitag_label_valid:** Valid or invalid tag (Valid/Invalid).
87
+ - **valid_comment:** Comment on the tag.
88
+ - **paragraphA_article:** Title of article containing passage 1.
89
+ - **paragraphA_information:** Relevant information of passage 1.
90
+ - **paragraphA_information_standalone:** Decontextualized relevant information of passage 1.
91
+ - **paragraphB_article:** Relevant information of passage 2.
92
+ - **paragraphB_information_standalone:** Decontextualized relevant information of passage 2.
93
+ - **wikitag_label_samepassage:** Boolean value stating whether passage 1 and passage 2 are the same (Same/Different).
94
+ - **relevantInfo_comment_A:** Comment on the information of passage 1.
95
+ - **relevantInfo_comment_B:** Comment on the information of passage 2.
96
+ - **Contradict type I:** Contradiction type I focuses on the fine-grained semantics of the contradiction, e.g., date/time, location, language, etc.
97
+ - **Contradict type II:** Contradiction type II focuses on the modality the contradiction. It describes the modality of passage 1 and passage 2, whether the information is from a piece of text, or from a row an infobox or a table.
98
+ - **Contradict type III:** Contradiction type III focuses on the source the contradiction. It describes whether passage 1 and passage 2 are from the same article or not.
99
+ - **Contradict type IV:** Contradiction type IV focuses on the reasoning aspect. It describes whether the contraction is explicit or implicit (Explicit/Implicit). Implicit contradiction requires some reasoning to understand why passage 1 and passage 2 are contradicted.
100
+ - **question1:** Question 1 inferred from the contradiction.
101
+ - **question1_answer1:** Gold answer to question 1 according to passage 1.
102
+ - **question1_answer2:** Gold answer to question 1 according to passage 2.
103
+ - **question2:** Question 2 inferred from the contradiction.
104
+ - **question2_answer1:** Gold answer to question 2 according to passage 1.
105
+ - **question2_answer2:** Gold answer to question 2 according to passage 2.
 
 
106
 
107
 
108
  ## Usage of the Dataset
109
 
110
+ We provide the following starter code. Please refer to the [GitHub repository](https://github.com/) for more information about the functions ```load_testingdata``` and ```generateAnswers_bam_models```.
111
 
112
 
113
  ```python
 
146
  max_new_tokens=250,
147
  min_new_tokens=1,
148
  decoding_method=DecodingMethod.GREEDY,
 
149
  return_options=TextGenerationReturnOptions(
150
  # if ordered is False, you can use return_options to retrieve the corresponding prompt
151
  input_text=True,
152
  ),
153
  )
154
 
155
+
156
+ # load dataset
157
  testingUnits = load_testingdata()
158
+ # test LLMs models
159
  generateAnswers_bam_models(testingUnits)
160
  ```
161
 
 
245
 
246
  **APA:**
247
 
248
+ Hou, Y., Pascale, A., Carnerero-Cano, J., Tchrakian, T., Marinescu, R., Daly, E., Padhi, I., & Sattigeri, P. (2024). WikiContradict: A Benchmark for Evaluating LLMs on Real-World Knowledge Conflicts from Wikipedia. arXiv preprint arXiv:2406.13805.
249
 
250
  <!-- ## Glossary [optional] -->
251