Add Open-QA training annotations
Browse files- README.md +10 -1
- SHA256SUMS +3 -2
- annotations/train.open_qa.jsonl.gz +3 -0
- dataset_info.json +18 -4
README.md
CHANGED
|
@@ -2,7 +2,16 @@
|
|
| 2 |
|
| 3 |
Training dataset for **VisualWorldBench: A Fine-Grained Multi-Task Benchmark for Evaluating Visual World Knowledge in MLLMs**.
|
| 4 |
|
| 5 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
|
| 7 |
Training images are provided through upstream links. Obtain them under the original providers' access conditions and replace `images` with local file paths before training.
|
| 8 |
|
|
|
|
| 2 |
|
| 3 |
Training dataset for **VisualWorldBench: A Fine-Grained Multi-Task Benchmark for Evaluating Visual World Knowledge in MLLMs**.
|
| 4 |
|
| 5 |
+
The release contains **7,235,746 training records** across MCQ and Open-QA, covering **53,999 distinct target names**.
|
| 6 |
+
|
| 7 |
+
| Task | Complete annotation file | Records |
|
| 8 |
+
|---|---|---:|
|
| 9 |
+
| MCQ | `annotations/train.mcq.v3.jsonl.gz` | 6,850,706 |
|
| 10 |
+
| Open-QA | `annotations/train.open_qa.jsonl.gz` | 385,040 |
|
| 11 |
+
|
| 12 |
+
Each compressed JSONL file contains one training record per line. Each record contains `messages` and `images`. The `images` field provides a direct image URL or an official dataset download page. Records with dataset-page URLs also include `path`, which locates the image within the extracted dataset.
|
| 13 |
+
|
| 14 |
+
MCQ includes both ordinary and reasoning responses. Open-QA responses use `<think>...</think>` and `<answer>...</answer>`.
|
| 15 |
|
| 16 |
Training images are provided through upstream links. Obtain them under the original providers' access conditions and replace `images` with local file paths before training.
|
| 17 |
|
SHA256SUMS
CHANGED
|
@@ -1,4 +1,4 @@
|
|
| 1 |
-
|
| 2 |
2d1d0ec253d55a3f63e6d649950568d9590855e12489ac699bf47e7a5a0ca8de annotations/train.mcq.v3.jsonl.gz
|
| 3 |
e4f262c4d094c009f343f958976f3849ebfc673dd335957f8debb6a27b922996 annotations/train.mcq.v3.part-00000.jsonl.gz
|
| 4 |
dc5fb21c07064ea53088ac386f7ec4696b89eed89778b4ff190de00b0acedaef annotations/train.mcq.v3.part-00001.jsonl.gz
|
|
@@ -47,7 +47,8 @@ d51e4eef6ecc4d9f9fce40bb63326c253c3ff083a1add58a01b8603172f2ac0b annotations/tr
|
|
| 47 |
7ce719a3d19e98ff5d8dbbff5dbbc39a6aaa7a3712b47b2ae1d488251c601b76 annotations/train.mcq.v3.part-00044.jsonl.gz
|
| 48 |
77ded025d9be9790cd4b639b048f6c61fb6a6123f245012ab0e1303b4a48a16f annotations/train.mcq.v3.part-00045.jsonl.gz
|
| 49 |
3ab2c9019dda7da64c0a8e5be0c2492533bb11b2e2e5c9d720e94c7712183a57 annotations/train.mcq.v3.part-00046.jsonl.gz
|
| 50 |
-
|
|
|
|
| 51 |
90fe6df45bd6040a5e20a084fb80e81c0d4747abfca18f2227b46fa6d9845a13 image_shards/part-00000.manifest.json
|
| 52 |
c78a7ccd7b4c42549adb260a20bdf0c46241653ed7b293b6ca54300114f4bb34 image_shards/part-00001.manifest.json
|
| 53 |
c21a2f98927a6852caf31d6508ec31169389f22096c1028014f30dfbf351e9db image_shards/part-00002.manifest.json
|
|
|
|
| 1 |
+
6b53e62212d1ba481b9dae8fb06b336c51550bef6af7f2c790963afa2d13dcd6 README.md
|
| 2 |
2d1d0ec253d55a3f63e6d649950568d9590855e12489ac699bf47e7a5a0ca8de annotations/train.mcq.v3.jsonl.gz
|
| 3 |
e4f262c4d094c009f343f958976f3849ebfc673dd335957f8debb6a27b922996 annotations/train.mcq.v3.part-00000.jsonl.gz
|
| 4 |
dc5fb21c07064ea53088ac386f7ec4696b89eed89778b4ff190de00b0acedaef annotations/train.mcq.v3.part-00001.jsonl.gz
|
|
|
|
| 47 |
7ce719a3d19e98ff5d8dbbff5dbbc39a6aaa7a3712b47b2ae1d488251c601b76 annotations/train.mcq.v3.part-00044.jsonl.gz
|
| 48 |
77ded025d9be9790cd4b639b048f6c61fb6a6123f245012ab0e1303b4a48a16f annotations/train.mcq.v3.part-00045.jsonl.gz
|
| 49 |
3ab2c9019dda7da64c0a8e5be0c2492533bb11b2e2e5c9d720e94c7712183a57 annotations/train.mcq.v3.part-00046.jsonl.gz
|
| 50 |
+
a5f1e1fe4cd882ef2bb0e30ae3696da75d7a41829a644284384c3554d9de816b annotations/train.open_qa.jsonl.gz
|
| 51 |
+
7f650a5fa703862f0bfb61b359d5110cd804e382f8ee71a8ce37558d64c8f341 dataset_info.json
|
| 52 |
90fe6df45bd6040a5e20a084fb80e81c0d4747abfca18f2227b46fa6d9845a13 image_shards/part-00000.manifest.json
|
| 53 |
c78a7ccd7b4c42549adb260a20bdf0c46241653ed7b293b6ca54300114f4bb34 image_shards/part-00001.manifest.json
|
| 54 |
c21a2f98927a6852caf31d6508ec31169389f22096c1028014f30dfbf351e9db image_shards/part-00002.manifest.json
|
annotations/train.open_qa.jsonl.gz
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a5f1e1fe4cd882ef2bb0e30ae3696da75d7a41829a644284384c3554d9de816b
|
| 3 |
+
size 80634128
|
dataset_info.json
CHANGED
|
@@ -1,12 +1,12 @@
|
|
| 1 |
{
|
| 2 |
-
"records":
|
| 3 |
"plain_mcq_records": 6254039,
|
| 4 |
"reasoning_mcq_records": 596667,
|
| 5 |
-
"unique_target_names_exact":
|
| 6 |
"image_paths": "images contains upstream HTTP(S) URLs: direct image links or official dataset acquisition pages. For dataset-page records, path identifies the image relative to the extracted source dataset root.",
|
| 7 |
"image_reference_types": {
|
| 8 |
-
"dataset_page":
|
| 9 |
-
"direct_image":
|
| 10 |
},
|
| 11 |
"image_distribution": "upstream_links",
|
| 12 |
"hosted_image_files": 0,
|
|
@@ -16,5 +16,19 @@
|
|
| 16 |
],
|
| 17 |
"optional_record_fields": {
|
| 18 |
"path": "Required for dataset-page records; relative image path within the extracted source dataset."
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
}
|
| 20 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"records": 7235746,
|
| 3 |
"plain_mcq_records": 6254039,
|
| 4 |
"reasoning_mcq_records": 596667,
|
| 5 |
+
"unique_target_names_exact": 53999,
|
| 6 |
"image_paths": "images contains upstream HTTP(S) URLs: direct image links or official dataset acquisition pages. For dataset-page records, path identifies the image relative to the extracted source dataset root.",
|
| 7 |
"image_reference_types": {
|
| 8 |
+
"dataset_page": 1616480,
|
| 9 |
+
"direct_image": 5619266
|
| 10 |
},
|
| 11 |
"image_distribution": "upstream_links",
|
| 12 |
"hosted_image_files": 0,
|
|
|
|
| 16 |
],
|
| 17 |
"optional_record_fields": {
|
| 18 |
"path": "Required for dataset-page records; relative image path within the extracted source dataset."
|
| 19 |
+
},
|
| 20 |
+
"mcq_records": 6850706,
|
| 21 |
+
"open_qa_records": 385040,
|
| 22 |
+
"annotation_files": {
|
| 23 |
+
"mcq": {
|
| 24 |
+
"file": "annotations/train.mcq.v3.jsonl.gz",
|
| 25 |
+
"records": 6850706,
|
| 26 |
+
"unique_target_names_exact": 48063
|
| 27 |
+
},
|
| 28 |
+
"open_qa": {
|
| 29 |
+
"file": "annotations/train.open_qa.jsonl.gz",
|
| 30 |
+
"records": 385040,
|
| 31 |
+
"unique_target_names_exact": 30522
|
| 32 |
+
}
|
| 33 |
}
|
| 34 |
}
|