Dataset Viewer
Auto-converted to Parquet Duplicate
id
stringlengths
1
4
question_id
stringlengths
1
4
question
stringclasses
70 values
answer
stringclasses
2 values
image_source
stringlengths
25
25
image
imagewidth (px)
333
640
category
stringclasses
3 values
7
8
Is there a car in the image?
no
COCO_val2014_000000210789
adversarial
15
16
Is there a motorcycle in the image?
no
COCO_val2014_000000429109
adversarial
17
18
Is there a truck in the image?
no
COCO_val2014_000000429109
adversarial
23
24
Is there a truck in the image?
no
COCO_val2014_000000211674
adversarial
25
26
Is there a bus in the image?
no
COCO_val2014_000000458338
adversarial
29
30
Is there a truck in the image?
no
COCO_val2014_000000458338
adversarial
33
34
Is there a chair in the image?
no
COCO_val2014_000000283412
adversarial
35
36
Is there a book in the image?
no
COCO_val2014_000000283412
adversarial
39
40
Is there a cup in the image?
no
COCO_val2014_000000265719
adversarial
41
42
Is there a dining table in the image?
no
COCO_val2014_000000265719
adversarial
42
43
Is there a tv in the image?
yes
COCO_val2014_000000461331
adversarial
49
50
Is there a handbag in the image?
no
COCO_val2014_000000544456
adversarial
53
54
Is there a snowboard in the image?
no
COCO_val2014_000000544456
adversarial
67
68
Is there a handbag in the image?
no
COCO_val2014_000000353180
adversarial
69
70
Is there a car in the image?
no
COCO_val2014_000000353180
adversarial
71
72
Is there a traffic light in the image?
no
COCO_val2014_000000353180
adversarial
81
82
Is there a bowl in the image?
no
COCO_val2014_000000569839
adversarial
83
84
Is there a handbag in the image?
no
COCO_val2014_000000569839
adversarial
87
88
Is there a truck in the image?
no
COCO_val2014_000000219622
adversarial
101
102
Is there a backpack in the image?
no
COCO_val2014_000000482476
adversarial
103
104
Is there a sports ball in the image?
no
COCO_val2014_000000131115
adversarial
107
108
Is there a bench in the image?
no
COCO_val2014_000000131115
adversarial
109
110
Is there a bottle in the image?
no
COCO_val2014_000000157084
adversarial
120
121
Is there a bowl in the image?
yes
COCO_val2014_000000336872
adversarial
122
123
Is there a spoon in the image?
yes
COCO_val2014_000000336872
adversarial
129
130
Is there a couch in the image?
no
COCO_val2014_000000075591
adversarial
135
136
Is there a book in the image?
no
COCO_val2014_000000516916
adversarial
137
138
Is there a bottle in the image?
no
COCO_val2014_000000516916
adversarial
141
142
Is there a cup in the image?
no
COCO_val2014_000000542145
adversarial
148
149
Is there a car in the image?
yes
COCO_val2014_000000218224
adversarial
149
150
Is there a bus in the image?
no
COCO_val2014_000000218224
adversarial
151
152
Is there a handbag in the image?
no
COCO_val2014_000000297078
adversarial
155
156
Is there a snowboard in the image?
no
COCO_val2014_000000297078
adversarial
157
158
Is there a dining table in the image?
no
COCO_val2014_000000033270
adversarial
158
159
Is there a person in the image?
yes
COCO_val2014_000000033270
adversarial
161
162
Is there a chair in the image?
no
COCO_val2014_000000033270
adversarial
163
164
Is there a car in the image?
no
COCO_val2014_000000140583
adversarial
171
172
Is there a traffic light in the image?
no
COCO_val2014_000000421455
adversarial
176
177
Is there a toothbrush in the image?
yes
COCO_val2014_000000288639
adversarial
177
178
Is there a cup in the image?
no
COCO_val2014_000000288639
adversarial
183
184
Is there a motorcycle in the image?
no
COCO_val2014_000000291936
adversarial
185
186
Is there a handbag in the image?
no
COCO_val2014_000000291936
adversarial
191
192
Is there a bench in the image?
no
COCO_val2014_000000063953
adversarial
196
197
Is there a dining table in the image?
yes
COCO_val2014_000000526321
adversarial
197
198
Is there a cup in the image?
no
COCO_val2014_000000526321
adversarial
199
200
Is there a skis in the image?
no
COCO_val2014_000000042190
adversarial
211
212
Is there a truck in the image?
no
COCO_val2014_000000170517
adversarial
216
217
Is there a person in the image?
yes
COCO_val2014_000000498759
adversarial
219
220
Is there a bicycle in the image?
no
COCO_val2014_000000498759
adversarial
221
222
Is there a truck in the image?
no
COCO_val2014_000000498759
adversarial
225
226
Is there a dining table in the image?
no
COCO_val2014_000000360600
adversarial
227
228
Is there a spoon in the image?
no
COCO_val2014_000000360600
adversarial
228
229
Is there a tv in the image?
yes
COCO_val2014_000000031773
adversarial
229
230
Is there a chair in the image?
no
COCO_val2014_000000031773
adversarial
233
234
Is there a cell phone in the image?
no
COCO_val2014_000000031773
adversarial
237
238
Is there a handbag in the image?
no
COCO_val2014_000000500257
adversarial
240
241
Is there a car in the image?
yes
COCO_val2014_000000574057
adversarial
247
248
Is there a bus in the image?
no
COCO_val2014_000000456178
adversarial
255
256
Is there a spoon in the image?
no
COCO_val2014_000000565941
adversarial
265
266
Is there a traffic light in the image?
no
COCO_val2014_000000454642
adversarial
269
270
Is there a bicycle in the image?
no
COCO_val2014_000000454642
adversarial
271
272
Is there a handbag in the image?
no
COCO_val2014_000000205729
adversarial
275
276
Is there a snowboard in the image?
no
COCO_val2014_000000205729
adversarial
285
286
Is there a backpack in the image?
no
COCO_val2014_000000329717
adversarial
288
289
Is there a chair in the image?
yes
COCO_val2014_000000012333
adversarial
297
298
Is there a cup in the image?
no
COCO_val2014_000000480122
adversarial
301
302
Is there a dining table in the image?
no
COCO_val2014_000000515904
adversarial
303
304
Is there a bottle in the image?
no
COCO_val2014_000000515904
adversarial
307
308
Is there a baseball bat in the image?
no
COCO_val2014_000000437347
adversarial
311
312
Is there a tennis racket in the image?
no
COCO_val2014_000000437347
adversarial
315
316
Is there a truck in the image?
no
COCO_val2014_000000354229
adversarial
317
318
Is there a traffic light in the image?
no
COCO_val2014_000000354229
adversarial
333
334
Is there a cup in the image?
no
COCO_val2014_000000217397
adversarial
343
344
Is there a truck in the image?
no
COCO_val2014_000000054025
adversarial
349
350
Is there a car in the image?
no
COCO_val2014_000000084447
adversarial
354
355
Is there a person in the image?
yes
COCO_val2014_000000192660
adversarial
357
358
Is there a cup in the image?
no
COCO_val2014_000000192660
adversarial
359
360
Is there a bottle in the image?
no
COCO_val2014_000000192660
adversarial
361
362
Is there a cup in the image?
no
COCO_val2014_000000575755
adversarial
363
364
Is there a bowl in the image?
no
COCO_val2014_000000575755
adversarial
367
368
Is there a chair in the image?
no
COCO_val2014_000000354088
adversarial
369
370
Is there a truck in the image?
no
COCO_val2014_000000354088
adversarial
371
372
Is there a cup in the image?
no
COCO_val2014_000000354088
adversarial
376
377
Is there a handbag in the image?
yes
COCO_val2014_000000311327
adversarial
377
378
Is there a backpack in the image?
no
COCO_val2014_000000311327
adversarial
385
386
Is there a bowl in the image?
no
COCO_val2014_000000350898
adversarial
395
396
Is there a vase in the image?
no
COCO_val2014_000000170365
adversarial
398
399
Is there a clock in the image?
yes
COCO_val2014_000000021645
adversarial
405
406
Is there a truck in the image?
no
COCO_val2014_000000528905
adversarial
413
414
Is there a dining table in the image?
no
COCO_val2014_000000239347
adversarial
415
416
Is there a chair in the image?
no
COCO_val2014_000000007320
adversarial
419
420
Is there a couch in the image?
no
COCO_val2014_000000007320
adversarial
433
434
Is there a handbag in the image?
no
COCO_val2014_000000564336
adversarial
435
436
Is there a traffic light in the image?
no
COCO_val2014_000000564336
adversarial
447
448
Is there a chair in the image?
no
COCO_val2014_000000465275
adversarial
449
450
Is there a tv in the image?
no
COCO_val2014_000000465275
adversarial
479
480
Is there a handbag in the image?
no
COCO_val2014_000000332025
adversarial
483
484
Is there a dog in the image?
no
COCO_val2014_000000318204
adversarial
497
498
Is there a backpack in the image?
no
COCO_val2014_000000069189
adversarial
499
500
Is there a skis in the image?
no
COCO_val2014_000000199764
adversarial
End of preview. Expand in Data Studio

lmms-lab_POPE-problematic

🔥🔥 Update: Seems like some other researchers have already figured out this: https://arxiv.org/abs/2504.15707v1. They also have published corrected annotations at: https://github.com/YanNeu/RePOPE/tree/main/annotations. See https://huggingface.co/datasets/SushantGautam/RePOPE for the corrected version

This dataset contains potentially incorrect annotations from the POPE dataset that were automatically flagged during experimentation with vision-language models.

The goal of this dataset is community review and verification. The examples included here are suspected to have incorrect answers, but they are not guaranteed to be wrong. Contributors and researchers are encouraged to inspect them and determine whether the original annotation is indeed incorrect.


Motivation

While working with the original dataset:

https://huggingface.co/datasets/lmms-lab/POPE

I noticed that some examples appear to have incorrect ground-truth answers when visually inspected.

To investigate this further, I ran a filtering pipeline to detect potential annotation mismatches between the image and the labeled answer. The filtered samples are collected in this dataset for manual inspection and discussion.

The intention is not to replace the original dataset, but to highlight examples that may require re-evaluation or correction.


How the samples were filtered

The suspected problematic samples were identified using:

Model: Qwen2-VL

Method:
The model was prompted with the image and the POPE question. If the model's prediction strongly disagreed with the labeled answer, the sample was flagged as potentially problematic.

Important notes:

  • This filtering was performed as part of a separate experiment on POPE.
  • Qwen2-VL is not treated as a ground truth verifier.
  • Some flagged samples may still be correctly labeled.

Future verification with stronger models or human review may help determine the true correctness.


Intended Use

This dataset is meant for:

  • Manual dataset auditing
  • Community review
  • Benchmark quality analysis
  • Studying annotation errors in vision-language datasets

Possible workflows include:

  • Human verification of each flagged sample
  • Cross-model agreement analysis
  • Dataset cleaning experiments
  • Robustness evaluation of VLM hallucination benchmarks

Dataset Structure

The dataset contains a subset of samples from the original POPE dataset that were flagged as suspicious.

Each example includes fields similar to the original dataset:

  • id
  • question_id
  • question
  • answer
  • image_source
  • image
  • category

These correspond directly to entries in the original dataset.


Important Disclaimer

⚠️ These samples are only suspected to be problematic.

They were filtered automatically and may include:

  • genuine annotation errors
  • model mistakes
  • ambiguous images
  • borderline cases

Human verification is required before any conclusions are drawn.

Downloads last month
11

Paper for SushantGautam/lmms-lab_POPE-problematic