zt-circlecat commited on
Commit
2279d02
·
verified ·
1 Parent(s): 47e9ca5

Add PyMETA dataset: splits + dataset card

Browse files
Files changed (5) hide show
  1. .gitattributes +3 -0
  2. README.md +152 -0
  3. dev.csv +3 -0
  4. test.csv +3 -0
  5. train.csv +3 -0
.gitattributes CHANGED
@@ -58,3 +58,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
58
  # Video files - compressed
59
  *.mp4 filter=lfs diff=lfs merge=lfs -text
60
  *.webm filter=lfs diff=lfs merge=lfs -text
 
 
 
 
58
  # Video files - compressed
59
  *.mp4 filter=lfs diff=lfs merge=lfs -text
60
  *.webm filter=lfs diff=lfs merge=lfs -text
61
+ dev.csv filter=lfs diff=lfs merge=lfs -text
62
+ test.csv filter=lfs diff=lfs merge=lfs -text
63
+ train.csv filter=lfs diff=lfs merge=lfs -text
README.md CHANGED
@@ -1,3 +1,155 @@
1
  ---
 
2
  license: cc-by-nc-4.0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ pretty_name: PyMETA
3
  license: cc-by-nc-4.0
4
+ language:
5
+ - en
6
+ - zh
7
+ annotations_creators:
8
+ - expert-generated
9
+ - machine-generated
10
+ language_creators:
11
+ - found
12
+ multilinguality:
13
+ - monolingual
14
+ size_categories:
15
+ - 10K<n<100K
16
+ source_datasets:
17
+ - original
18
+ task_categories:
19
+ - text-classification
20
+ task_ids:
21
+ - multi-class-classification
22
+ - multi-label-classification
23
+ tags:
24
+ - code
25
+ - python
26
+ - code-error-classification
27
+ - education
28
+ - online-judge
29
+ configs:
30
+ - config_name: default
31
+ data_files:
32
+ - split: train
33
+ path: train.csv
34
+ - split: validation
35
+ path: dev.csv
36
+ - split: test
37
+ path: test.csv
38
  ---
39
+
40
+ # Dataset Card for PyMETA
41
+
42
+ ## Dataset Description
43
+
44
+ - **Repository (code, prompts, scripts):** https://github.com/Circle-Cat/pymeta
45
+ - **Paper:** *PyMETA: A Benchmark Dataset for Hierarchical Student Code Error Classification with Python-Interpreter-Based Labels*
46
+ - **Point of Contact:** CircleCat (`{cyli, ztang}@circlecat.org`)
47
+
48
+ ### Dataset Summary
49
+
50
+ **PyMETA** (**Py**thon **M**ulti-**E**rror **TA**xonomy) is a large-scale benchmark for
51
+ **hierarchical student code error classification**. It contains **48,646** real Python
52
+ code submissions from **579** users across **155** distinct problems (22 problem types),
53
+ collected from the Circle Cat online learning platform. Every submission has a
54
+ **single-error label** derived from Online Judge execution output, organized under a
55
+ **three-level hierarchical taxonomy** (binary → three-class → 14 fine-grained error types
56
+ grounded in Python's official exception hierarchy).
57
+
58
+ An **expert-annotated multi-error diagnostic subset** of 97 samples (13 error types) is
59
+ provided in the [GitHub repository](https://github.com/Circle-Cat/pymeta) for studying
60
+ co-occurring errors.
61
+
62
+ ### Supported Tasks
63
+
64
+ - **Task A — Binary classification:** `No Error` vs. `Error`.
65
+ - **Task B — Three-class classification:** `No Error` / `Logic Error` / `Explicit Error`.
66
+ - **Task C — Multi-class classification:** 14 fine-grained error types.
67
+ - **Multi-error classification** (on the 97-sample subset): identify *all* concurrent
68
+ errors in a submission.
69
+
70
+ ### Languages
71
+
72
+ Student code is **Python**. Problem descriptions are primarily in **Chinese** (`zh`); the
73
+ evaluation prompts are in **English** (`en`).
74
+
75
+ ## How to use
76
+
77
+ ```python
78
+ from datasets import load_dataset
79
+
80
+ ds = load_dataset("CircleCat/pymeta")
81
+ print(ds["train"][0])
82
+ ```
83
+
84
+ ## Dataset Structure
85
+
86
+ ### Data Fields
87
+
88
+ Each sample has 9 features plus a single-error label:
89
+
90
+ | Field | Type | Description |
91
+ |-------|------|-------------|
92
+ | `userId` | string | Anonymized numeric identifier of the student |
93
+ | `name` | string | Problem/lesson name (problem type) |
94
+ | `questionId` | string | Unique identifier of the problem |
95
+ | `question` | string | Problem description |
96
+ | `exceptedAnswer` | string | A correct reference code solution |
97
+ | `attemptId` | string | Attempt number |
98
+ | `studentAnswer` | string | The student's submitted Python code |
99
+ | `testOutcome` | string | Online Judge error message / execution output |
100
+ | `attemptstepid` | string | Step identifier for the attempt |
101
+ | `error_category` | string | Single-error label (see taxonomy below) |
102
+
103
+ ### Taxonomy (label IDs)
104
+
105
+ `0` No Error · `1` Logic Error · `2` Syntax Error · `3` Name Error · `4` Type Error ·
106
+ `5` Indentation Error · `6` Unbound Local Error · `7` Key Error · `8` Index Error ·
107
+ `9` EOF Error · `10` Recursion Error · `11` Value Error · `12` Tab Error ·
108
+ `13` Other Errors. Full definitions and per-class counts are in the
109
+ [GitHub `TAXONOMY.md`](https://github.com/Circle-Cat/pymeta/blob/main/TAXONOMY.md).
110
+
111
+ ### Data Splits
112
+
113
+ | Split | Examples |
114
+ |-------|---------:|
115
+ | train | 39,402 |
116
+ | validation | 4,379 |
117
+ | test | 4,865 |
118
+ | **total** | **48,646** |
119
+
120
+ ## Source Data
121
+
122
+ Submissions were collected from historical logs of the Circle Cat online learning
123
+ platform (a self-hosted Moodle instance with an integrated Online Judge). They were
124
+ generated organically by learners of varying proficiency during ordinary coursework, so
125
+ they provide a realistic and diverse distribution of student code errors.
126
+
127
+ ## Personal and Sensitive Information
128
+
129
+ The data were collected from pre-existing educational records under the platform's terms
130
+ of use (which inform users that anonymized data may be used for educational and research
131
+ purposes), with no recruitment or experimental intervention. **All identifiers have been
132
+ irreversibly anonymized**; the dataset contains **no personally identifiable information
133
+ (PII)** and no offensive content.
134
+
135
+ ## Licensing Information
136
+
137
+ Released under the
138
+ [Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0)](https://creativecommons.org/licenses/by-nc/4.0/)
139
+ license.
140
+
141
+ ## Citation
142
+
143
+ ```bibtex
144
+ @misc{li2026pymeta,
145
+ title = {PyMETA: A Benchmark Dataset for Hierarchical Student Code Error
146
+ Classification with Python-Interpreter-Based Labels},
147
+ author = {Li, Chuyue and Tang, Ziqi and Wang, Jingyi and Wu, Yu and
148
+ Hashimoto, Kazuma and Gao, Lingyu},
149
+ year = {2026},
150
+ howpublished = {\url{https://github.com/Circle-Cat/pymeta}},
151
+ note = {CircleCat}
152
+ }
153
+ ```
154
+
155
+ <!-- TODO: replace with the final arXiv / ACL citation once available. -->
dev.csv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:11e707c4d5c9134714d5b53abee8f651ec9bfb1da248bd9f27336689c63dc990
3
+ size 12350899
test.csv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7c2bc9701a4560303aa1d62583d5065291e88fa73df305ff54718aa33375b079
3
+ size 13607887
train.csv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9a48c2b3822e135200bd63b0eeaf3492027fdebb13e4a8fa9ef8964af51128d3
3
+ size 109652461