Datasets:

SI-Lab
/

StepCountQA-RL-Traj_11_50

dataset_info:
  features:
    - name: images
      sequence: image
    - name: problem
      dtype: string
    - name: answer
      dtype: string
  splits:
    - name: train
      num_bytes: 2401285909
      num_examples: 9508
  download_size: 0
  dataset_size: 2401285909
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

StepCountQA-RL-Traj_11_50 Dataset

Dataset Description

StepCountQA-RL-Traj_11_50 is a visual question answering dataset focused on counting and spatial reasoning tasks. The dataset contains images paired with questions and detailed answers that require understanding and analyzing visual content.

Dataset Structure

Data Fields

images: A sequence of images (typically one image per sample)
problem: The question text describing the task to be solved in the image
answer: The answer text containing the solution to the problem

Data Splits

train: 9508 training samples

Dataset Statistics

Training set size: 9508 samples
Count range: 11 - 50
Total dataset size: approximately 2290.0 MB

Count Distribution (11-50)

Count	Samples
11	707
12	851
13	517
14	534
15	498
16	441
17	318
18	390
19	262
20	389
21	279
22	246
23	201
24	288
25	239
26	220
27	191
28	203
29	141
30	202
31	161
32	173
33	138
34	141
35	142
36	166
37	105
38	115
39	104
40	170
41	93
42	110
43	91
44	109
45	98
46	77
47	65
48	126
49	85
50	122

Usage Example

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("SI-Lab/StepCountQA-RL-Traj_11_50")

# Access the training set
train_data = dataset["train"]

# View a sample
print(train_data[0])
# Output: {'images': [...], 'problem': '...', 'answer': '...'}

Dataset Creation

This dataset was created for training visual reasoning models, particularly focusing on: