Update README.md

README.md

@@ -1,10 +1,52 @@
----
-license: mit
-task_categories:
-- question-answering
-language:
-- en
-size_categories:
-- 1K<n<10K
----
-**MV-MATH** is a meticulously curated dataset of 2,009 high-quality mathematical problems. Each problem integrates multiple images interleaved with text, derived from authentic K-12 scenarios, and enriched with detailed annotations. **MV-MATH** includes multiple-choice, free-form, and multi-step questions, covering 11 subject areas across 3 difficulty levels, and serves as a comprehensive and rigorous benchmark for assessing MLLMs’ mathematical reasoning in multi-visual contexts.
+---
+license: mit
+task_categories:
+- question-answering
+language:
+- en
+size_categories:
+- 1K<n<10K
+---
+**Introduction of MV-MATH:**
+
+**MV-MATH** is a meticulously curated dataset of 2,009 high-quality mathematical problems. Each problem integrates multiple images interleaved with text, derived from authentic K-12 scenarios, and enriched with detailed annotations. **MV-MATH** includes multiple-choice, free-form, and multi-step questions, covering 11 subject areas across 3 difficulty levels, and serves as a comprehensive and rigorous benchmark for assessing MLLMs’ mathematical reasoning in multi-visual contexts.
+
+
+**Data Format**
+The dataset is provided in json format and contains the following attributes:
+```python
+{
+    "problem_id": the id number of each problem,
+    "question": the input question text,
+    "input_image": the images related to the question,
+    "is_multi_img": true, every question in MV-MATH is with more than 2 images,
+    "answer": the answer of the question,
+    "answer_type": choice or free-form or multi-step,
+    "difficulty": Low, Medium or High,
+    "grade": Elementary, Junior or Senior,
+    "subject": Analytic Geometry, Algebra, Metric Geometry, Combinatorics, Transformation Geometry, Logic, Solid Geometry, Arithmetic, Combinatorial Geometry, Descriptive Geometry or Statistics,
+    "image_relavance": 0 or 1, 1 represents that images are interrelated and understanding one image necessitates information from another, and 0 means that images are unrelated and can be interpreted independently without reference to other images，
+    "analysis": More detailed problem-solving process
+}
+```
+Here is an example:
+```python
+{
+    "problem_id": 1,
+    "question": "In the given solid shapes, which can be obtained by rotating a plane figure around a certain straight line for a full revolution?\nA.\n\n<image_1>\nB.\n\n<image_2>\nC.\n\n<image_3>\nD.\n\n<image_4>",
+    "input_image": [
+        "batch10-2024_06_14_13894f5e3499bf65de58g_0006_1.jpg",
+        "batch10-2024_06_14_13894f5e3499bf65de58g_0006_2.jpg",
+        "batch10-2024_06_14_13894f5e3499bf65de58g_0006_3.jpg",
+        "batch10-2024_06_14_13894f5e3499bf65de58g_0006_4.jpg"
+    ],
+    "is_multi_img": true,
+    "answer": "B",
+    "answer_type": "choice",
+    "difficulty": "Low",
+    "grade": "Junior",
+    "subject": "Solid Geometry",
+    "image_relavance": "0",
+    "analysis": "Solution: Since a solid obtained by rotating a plane figure around a certain straight line must have a curved surface,\n\nTherefore, option B fits the requirement;\n\nHence, the answer is: B.\n\n[Key Insight] This question examines the concepts of points, lines, surfaces, and solids. Understanding that \"a moving point forms a line, a moving line forms a surface, and a moving surface forms a solid\" is crucial to solving the problem."
+}
+```