Final validation of dataset
Browse files
README.md
CHANGED
@@ -14,7 +14,7 @@ This dataset thus contains 20 samples per sample type and programming language.
|
|
14 |
All data in this dataset are collected from 100 unique, open-source GitHub repositories with permissive licenses
|
15 |
that have >= 1000 stars, >= 5 branches, >= 10 contributors and are not a fork or archived. We collected the initial list of repositories using [SEART.](https://seart-ghs.si.usi.ch/)
|
16 |
|
17 |
-
Evaluation is to be performed by exact-match (EM) of diffs for the merge conflict setting and by LLM-as-a-Judge for the file-commit
|
18 |
|
19 |
# Supported Tasks
|
20 |
GitGoodBench Lite contains two types of samples: 'merge' and 'file_commit_chain'.
|
@@ -76,7 +76,6 @@ The following table provides per-field details. Columns marked “Yes” under *
|
|
76 |
| **scenario** | string | A JSON string describing specific scenario data (e.g., merge-conflict details, parent commits). | No |
|
77 |
| **sample_type** | string | The type of sample. Possible values: "merge" or "file_commit_chain." | No |
|
78 |
| **project_size** | string | Estimated size based on lines of code. Possible values: "tiny," "small," "medium," "large," or "huge." | Yes |
|
79 |
-
| **project_activity** | string | How recently the project was active. Possible values: "day," "week," or "month." | Yes |
|
80 |
| **difficulty** | string | The complexity level of the scenario. Possible values: "easy," "medium," or "hard." | Yes |
|
81 |
|
82 |
**Note**:
|
@@ -84,43 +83,37 @@ The following table provides per-field details. Columns marked “Yes” under *
|
|
84 |
- Fields marked **No** represent the primary data for the scenario. Use them to inform or categorize the scenario type and project details.
|
85 |
|
86 |
# Dataset statistics
|
87 |
-
We provide some statistics on the
|
88 |
|
89 |
## Dataset Skew
|
90 |
|
91 |
-
We note that our dataset is skewed towards the top three repositories especially, however skew flattens quickly.
|
92 |
|
93 |
### Distribution Statistics
|
94 |
-
- Total number of repositories (count):
|
95 |
-
- Average (mean) samples per repository: 1.
|
96 |
-
- Standard deviation (std): 0.
|
97 |
- Minimum (min): 1
|
98 |
- 25th percentile (25%): 1
|
99 |
- Median (50%): 1
|
100 |
- 75th percentile (75%): 1
|
101 |
-
- Maximum (max):
|
102 |
|
103 |
### Top-10 Repositories by Sample Count
|
104 |
|
105 |
| Repository | Percentage of Total Samples |
|
106 |
|------------------------------------------|----------------------------:|
|
107 |
-
| oss-review-toolkit/ort |
|
108 |
| stripe/stripe-android | 2.50% |
|
109 |
| element-hq/element-android | 2.50% |
|
110 |
-
| tlaplus/tlaplus | 1.67% |
|
111 |
| apache/hive | 1.67% |
|
|
|
112 |
| wikimedia/apps-android-wikipedia | 1.67% |
|
113 |
-
|
|
114 |
| liquibase/liquibase | 1.67% |
|
115 |
-
|
|
116 |
-
|
|
117 |
|
118 |
-
## Overall Difficulty Distribution
|
119 |
-
| Difficulty | Fraction |
|
120 |
-
|------------|----------|
|
121 |
-
| easy | 0.375000 |
|
122 |
-
| medium | 0.333333 |
|
123 |
-
| hard | 0.291667 |
|
124 |
|
125 |
## Difficulty Distribution for "merge" Scenarios
|
126 |
| Difficulty | Fraction |
|
@@ -129,12 +122,5 @@ We note that our dataset is skewed towards the top three repositories especially
|
|
129 |
| hard | 0.266667 |
|
130 |
| medium | 0.216667 |
|
131 |
|
132 |
-
## Difficulty Distribution for "file_commit_chain" Scenarios
|
133 |
-
| Difficulty | Fraction |
|
134 |
-
|------------|----------|
|
135 |
-
| medium | 0.450000 |
|
136 |
-
| hard | 0.316667 |
|
137 |
-
| easy | 0.233333 |
|
138 |
-
|
139 |
|
140 |
**Languages** We note that the text data in this dataset consists mostly of: commit messages, comments and is primarily in English. We do however not filter for any human languages explcitly.
|
|
|
14 |
All data in this dataset are collected from 100 unique, open-source GitHub repositories with permissive licenses
|
15 |
that have >= 1000 stars, >= 5 branches, >= 10 contributors and are not a fork or archived. We collected the initial list of repositories using [SEART.](https://seart-ghs.si.usi.ch/)
|
16 |
|
17 |
+
Evaluation is to be performed by exact-match (EM) of diffs for the merge conflict setting and by LLM-as-a-Judge for the file-commit chain setting. [For further details see our paper.]()
|
18 |
|
19 |
# Supported Tasks
|
20 |
GitGoodBench Lite contains two types of samples: 'merge' and 'file_commit_chain'.
|
|
|
76 |
| **scenario** | string | A JSON string describing specific scenario data (e.g., merge-conflict details, parent commits). | No |
|
77 |
| **sample_type** | string | The type of sample. Possible values: "merge" or "file_commit_chain." | No |
|
78 |
| **project_size** | string | Estimated size based on lines of code. Possible values: "tiny," "small," "medium," "large," or "huge." | Yes |
|
|
|
79 |
| **difficulty** | string | The complexity level of the scenario. Possible values: "easy," "medium," or "hard." | Yes |
|
80 |
|
81 |
**Note**:
|
|
|
83 |
- Fields marked **No** represent the primary data for the scenario. Use them to inform or categorize the scenario type and project details.
|
84 |
|
85 |
# Dataset statistics
|
86 |
+
We provide some statistics on the diversity of our dataset with respect to repositories and merge conflict resolution samples.
|
87 |
|
88 |
## Dataset Skew
|
89 |
|
90 |
+
We note that our dataset is skewed towards the top three repositories especially, however the skew flattens quickly.
|
91 |
|
92 |
### Distribution Statistics
|
93 |
+
- Total number of repositories (count): 100
|
94 |
+
- Average (mean) samples per repository: 1.2
|
95 |
+
- Standard deviation (std): 0.79
|
96 |
- Minimum (min): 1
|
97 |
- 25th percentile (25%): 1
|
98 |
- Median (50%): 1
|
99 |
- 75th percentile (75%): 1
|
100 |
+
- Maximum (max): 8
|
101 |
|
102 |
### Top-10 Repositories by Sample Count
|
103 |
|
104 |
| Repository | Percentage of Total Samples |
|
105 |
|------------------------------------------|----------------------------:|
|
106 |
+
| oss-review-toolkit/ort | 6.67% |
|
107 |
| stripe/stripe-android | 2.50% |
|
108 |
| element-hq/element-android | 2.50% |
|
|
|
109 |
| apache/hive | 1.67% |
|
110 |
+
| coil-kt/coil | 1.67% |
|
111 |
| wikimedia/apps-android-wikipedia | 1.67% |
|
112 |
+
| facebookresearch/habitat-lab | 1.67% |
|
113 |
| liquibase/liquibase | 1.67% |
|
114 |
+
| google/guava | 1.67% |
|
115 |
+
| kotlin/kotlinx.coroutines | 1.67% |
|
116 |
|
|
|
|
|
|
|
|
|
|
|
|
|
117 |
|
118 |
## Difficulty Distribution for "merge" Scenarios
|
119 |
| Difficulty | Fraction |
|
|
|
122 |
| hard | 0.266667 |
|
123 |
| medium | 0.216667 |
|
124 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
125 |
|
126 |
**Languages** We note that the text data in this dataset consists mostly of: commit messages, comments and is primarily in English. We do however not filter for any human languages explcitly.
|