LucasWeber commited on
Commit
6e34da2
•
1 Parent(s): 5cd73f9

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +97 -1
README.md CHANGED
@@ -6,4 +6,100 @@ language:
6
  pretty_name: The ICL consistency test
7
  size_categories:
8
  - 100K<n<1M
9
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6
  pretty_name: The ICL consistency test
7
  size_categories:
8
  - 100K<n<1M
9
+ ---
10
+ # The ICL consistency test
11
+
12
+ This 🤗 dataset provides data for the [GenBench CBT task 'The ICL consistency test'](https://github.com/GenBench/genbench_cbt/tree/main/src/genbench/tasks/icl_consistency_test).
13
+ The ICL consistency test measures the consistency of LLM predictions on the same data points across many different equivalent prompting setups.
14
+ The score in the associated metric (Cohen's kappa) can be understood as a measure of a model's prediction consistency in the face of task-irrelevant information.
15
+
16
+ For an easy evaluation of any 🤗 models, we refer to the code provided in the GenBench task. For in-depth information on the task, we refer to the associated
17
+ publications ([Weber et al., 2023](https://arxiv.org/abs/2312.04945),[2023](https://aclanthology.org/2023.conll-1.20/)) and the respective GenBench [doc.md](https://github.com/GenBench/genbench_cbt/blob/main/src/genbench/tasks/icl_consistency_test/doc.md).
18
+
19
+ Evaluation on the relevant metrics can be done via the _example_evaluation.py_ script in the [GenBench repository](https://github.com/GenBench/genbench_cbt/blob/main/src/genbench/tasks/icl_consistency_test/).
20
+
21
+ ### Dataset Description
22
+
23
+ _Abstract_: The ICL consistency test measures the consistency of LLM predictions on the same data points across many different prompting setups. Different setups are defined by "factors".
24
+ On the one hand, factors can be specific attributes of the used prompt (e.g. the number of examples the model is presented with ["n_shots"] or the type of instructions
25
+ that were used to wrap a specific datapoint ["Instructions"]). On the other hand, the analysis can also be augmented by factors that are related to the way a model is
26
+ evaluated (e.g. whether a model is calibrated) or the type of model that is evaluated (e.g. the number of parameters or instructions tuning). These external factors can
27
+ be added to the analysis by using the task.add_factor() method. The output metric is Cohen's kappa for each factor across all different conditions. A kappa value close to
28
+ 1 indicates that the factors do not change the model prediction, while a factor close to 0 strongly changes model predictions. The ICL consistency test has two subtasks,
29
+ one evaluating the ANLI-dataset ([Nie et al., 2019](https://aclanthology.org/N18-1101/)); the other the MNLI-dataset ([Wang et al., 2017](https://aclanthology.org/N18-1101/)).
30
+
31
+ _Size_: Each subtask contains 57600 when using the full 600 data_IDs. The user can choose to reduce the number of evaluated data_IDs.
32
+
33
+ - **Curated by:**
34
+ - resampling and arrangement was done by [Weber et al., 2023](https://arxiv.org/abs/2312.04945),[2023](https://aclanthology.org/2023.conll-1.20/);
35
+ - original data were curated by [Nie et al., 2019](https://aclanthology.org/N18-1101/) (ANLI) and [Wang et al., 2017](https://aclanthology.org/N18-1101/) (MNLI);
36
+ - templates were curated by [Bach et al., 2022](https://aclanthology.org/2022.acl-demo.9/) (promptsource).
37
+ - **Language:** English
38
+
39
+ ### Dataset Sources (basic links)
40
+
41
+ - **Repository:** Data files on [github](https://github.com/LucWeber/icl_consistency_data).
42
+ - **Paper:** [Weber et al., 2023](https://arxiv.org/abs/2312.04945),[2023](https://aclanthology.org/2023.conll-1.20/).
43
+ - **Demo:** Find pre-implemented code to evaluate any 🤗 model on [github](https://github.com/GenBench/genbench_cbt/blob/main/src/genbench/tasks/icl_consistency_test/example_evaluation.py).
44
+
45
+ ## Uses
46
+
47
+ In prompting, models are sensitive to task-irrelevant information in their prompt. This test can be used to quantify this sensitivity of any 🤗 model. The ICL consistency test does this by measuring a model's prediction consistency across many different semantically equivalent prompting setups.
48
+
49
+ ## Dataset Structure
50
+
51
+ <!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits,
52
+ relationships between data points, etc. -->
53
+
54
+ [_TBA_]
55
+
56
+ ## Dataset Creation
57
+
58
+ The data is a sample from the [MNLI](https://aclanthology.org/N18-1101/) and [ANLI](https://aclanthology.org/2020.acl-main.441/) datasets as well as prompt templates from [promptsource](https://aclanthology.org/2022.acl-demo.9/).
59
+ Please refer to the original publications's documentation for detailed information on dataset creation.
60
+
61
+ ## Bias, Risks, and Limitations
62
+
63
+ This dataset contains data from the [MNLI](https://aclanthology.org/N18-1101/) and [ANLI](https://aclanthology.org/2020.acl-main.441/) datasets and adheres to the same biases, risks and limitations.
64
+
65
+ ### Recommendations
66
+
67
+ We identify the following limitations of the consistency test:
68
+
69
+ 1. The number of factors is limited and does not cover all possible factors that might influence the predictions. We limited ourselves to factors we deem relevant, to ensure fast evaluation.
70
+
71
+ 2. Currently, the test is only implemented for the ANLI- and MNLI-datasets.
72
+
73
+ 3. Factors that are external to the dataset but should be considered in the analysis (e.g. _instruction tuning_ or _calibration_) have to be manually added by the user
74
+ using the task.add_factor() method (please use the GenBench implementation of the dataset. You can find it on [github](https://github.com/GenBench/genbench_cbt/tree/main/src/genbench/tasks/icl_consistency_test)).
75
+
76
+
77
+ ## Citation
78
+
79
+ This dataset was used in the following publications. If you use it, please consider citing the following references:
80
+
81
+ **BibTeX:**
82
+
83
+ ```
84
+ @inproceedings{weber2023mind,
85
+ title={Mind the instructions: a holistic evaluation of consistency and interactions in prompt-based learning},
86
+ author={Weber, Lucas and Bruni, Elia and Hupkes, Dieuwke},
87
+ booktitle={Proceedings of the 27th Conference on Computational Natural Language Learning (CoNLL)},
88
+ pages={294--313},
89
+ year={2023}
90
+ }
91
+ ```
92
+ ```
93
+ @article{weber2023icl,
94
+ title={The ICL Consistency Test},
95
+ author={Weber, Lucas and Bruni, Elia and Hupkes, Dieuwke},
96
+ journal={arXiv preprint arXiv:2312.04945},
97
+ year={2023}
98
+ }
99
+ ```
100
+
101
+ ## Dataset Card Authors
102
+ [Lucas Weber](https://lucweber.github.io/)
103
+
104
+ ## Dataset Card Contact
105
+ lucasweber000@gmail.com