Datasets:

Modalities:
Image
Text
Formats:
parquet
ArXiv:
Libraries:
Datasets
pandas
License:
yueqis commited on
Commit
06942b6
ยท
verified ยท
1 Parent(s): 01d4e9c

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +52 -0
README.md CHANGED
@@ -31,3 +31,55 @@ configs:
31
  - split: train
32
  path: data/train-*
33
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
31
  - split: train
32
  path: data/train-*
33
  ---
34
+
35
+ # VisualPuzzles: Decoupling Multimodal Reasoning Evaluation from Domain Knowledge
36
+
37
+ [๐Ÿ  Homepage](https://neulab.github.io/VisualPuzzles/) | [๐Ÿ“Š VisualPuzzles](https://huggingface.co/datasets/neulab/VisualPuzzles) | [๐Ÿ’ป Github](https://github.com/neulab/VisualPuzzles) | [๐Ÿ“„ Arxiv](https://arxiv.org/abs/2504.10342) | [๐Ÿ“• PDF](https://arxiv.org/pdf/2504.10342) | [๐Ÿ–ฅ๏ธ Zeno Model Output](https://hub.zenoml.com/project/2e727b03-a677-451a-b714-f2c07ad2b49f/VisualPuzzles)
38
+
39
+ ## Overview
40
+ **VisualPuzzles** is a multimodal benchmark specifically designed to evaluate **reasoning abilities** in large models while deliberately minimizing reliance on domain-specific knowledge.
41
+
42
+ Key features:
43
+ - 1168 diverse puzzles
44
+ - 5 reasoning categories: Algorithmic, Analogical, Deductive, Inductive, Spatial
45
+ - Difficulty labels: Easy, Medium, Hard
46
+ - Less knowledge-intensive than existing benchmarks (e.g., MMMU)
47
+ - More reasoning-complex than existing benchmarks (e.g., MMMU)
48
+
49
+ ## Key Findings
50
+ - All models perform worse than humans; most can't surpass even 5th-percentile human performance.
51
+ - Strong performance on knowledge-heavy benchmarks does not transfer well.
52
+ - Larger models and structured "thinking modes" don't guarantee better results.
53
+ - Scaling model size does not ensure stronger reasoning
54
+
55
+ ## Usage
56
+
57
+ To load this dataset via Hugging Faceโ€™s `datasets` library:
58
+
59
+ ```python
60
+ from datasets import load_dataset
61
+
62
+ dataset = load_dataset("neulab/VisualPuzzles")
63
+ data = dataset["train"]
64
+
65
+ sample = data[0]
66
+ print("ID:", sample["id"])
67
+ print("Category:", sample["category"])
68
+ print("Question:", sample["question"])
69
+ print("Options:", sample["options"])
70
+ print("Answer:", sample["answer"])
71
+ ```
72
+
73
+ ## Citation
74
+
75
+ If you use or reference this dataset in your work, please cite:
76
+
77
+ ```bibtex
78
+ @article{song2025visualpuzzles,
79
+ title = {VisualPuzzles: Decoupling Multimodal Reasoning Evaluation from Domain Knowledge},
80
+ author = {Song, Yueqi and Ou, Tianyue and Kong, Yibo and Li, Zecheng and Neubig, Graham and Yue, Xiang},
81
+ year = {2025},
82
+ journal = {arXiv preprint arXiv:2504.10342},
83
+ url = {https://arxiv.org/abs/2504.10342}
84
+ }
85
+ ```