Egemen Sert commited on
Commit
9b0cf30
·
verified ·
1 Parent(s): e2f40ae

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +66 -0
README.md CHANGED
@@ -24,4 +24,70 @@ configs:
24
  data_files:
25
  - split: train
26
  path: data/train-*
 
 
 
 
 
 
 
 
 
 
27
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
24
  data_files:
25
  - split: train
26
  path: data/train-*
27
+ license: mit
28
+ task_categories:
29
+ - image-text-to-text
30
+ - image-to-text
31
+ language:
32
+ - tr
33
+ - en
34
+ pretty_name: YKS - Uniform
35
+ size_categories:
36
+ - 1K<n<10K
37
  ---
38
+ # Dataset Card for Dataset Name
39
+
40
+ <!-- Provide a quick summary of the dataset. -->
41
+
42
+ Large Language Models (LLMs) and Large Multimodal Models (LMMs) demonstrate impressive problem-solving skills across many tasks and domains. However, their ability to reason over structured, curriculum-based educational questions—particularly in the context of Turkish high school entrance examinations—has not been systematically studied.
43
+
44
+ To address this gap, we introduce YKS Uniform, a balanced multimodal benchmark covering the Turkish high school curriculum with equal representation across all topics. By sampling six questions per topic, we constructed a dataset of 1,854 multimodal questions spanning both TYT and AYT exams. These questions require deep reasoning over text, diagrams, and exam-style contexts.
45
+
46
+ Using this benchmark, we conducted a comprehensive evaluation of 16 open-weight and proprietary models. Our results highlight both the strengths and limitations of current models in handling exam-style reasoning tasks. The best-performing system, Gemini-2.5-Flash, achieved an overall accuracy of %84.7, substantially higher than open-weight alternatives but still leaving a measurable gap to human-level performance.
47
+
48
+ Visit [YKS Uniform](https://yks-uniform.github.io/) webpage for details.
49
+
50
+ ## Dataset Details
51
+
52
+ ### Dataset Description
53
+
54
+ <!-- Provide a longer summary of what this dataset is. -->
55
+ 1854 multimodal exam preparation questions spanning 309 topics in Turkish High School curriculum. Each topic is equally represented by sampling first six questions of each topic's test from publicly available e-books of a publisher. For academic use only.
56
+
57
+
58
+ - **Curated by:** Egemen Sert, Şeyda Ertekin
59
+ - **Language(s) (NLP):** Turkish, English
60
+ - **License:** MIT
61
+
62
+ ### Dataset Sources [optional]
63
+
64
+ <!-- Provide the basic links for the dataset. -->
65
+
66
+ - **Repository:** [[More Information Needed]](https://yks-uniform.github.io/)
67
+ - **Paper [optional]:** [More Information Needed]
68
+ - **Demo [optional]:** [More Information Needed]
69
+
70
+ ## Uses
71
+
72
+ <!-- Address questions around how the dataset is intended to be used. -->
73
+
74
+ This is a benchmark dataset for academic use only. Please use the dataset exclusively for benchmarking purposes.
75
+
76
+ ## Citation [optional]
77
+
78
+ <!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->
79
+
80
+ **BibTeX:**
81
+
82
+ @misc{yksuniform2025,
83
+ title = {YKS Uniform: A Balanced Multimodal Benchmark Covering the Turkish High School Curriculum},
84
+ author = {Sert, Egemen and Ertekin, Şeyda},
85
+ year = {2025},
86
+ howpublished = {\url{https://yks-uniform.github.io/}},
87
+ note = {Accessed: 2025-08-23}
88
+ }
89
+
90
+
91
+ ## Dataset Card Contact
92
+
93
+ Please contact egemen.sert@metu.edu.tr for submissions and benchmark related questions.