Update README.md

README.md

@@ -7,21 +7,91 @@ dataset_info:
     dtype: string
   - name: answer
     dtype: string
-
   - name: problem_image_0
     dtype: image
-  - name: reasoning_image_0
-    dtype: image
-
   - name: resoning_thought_0
     dtype: string
+  - name: reasoning_image_0
+    dtype: image
   - name: resoning_thought_1
     dtype: string
+  - name: task
+    dtype: string
   - name: full_text_only_thought
     dtype: string
+  - name: meta_info
+    struct:
+    - name: map
+      sequence: 
+        sequence: int64
+    - name: icon_info
+      dtype: string
+    - name: task_type
+      dtype: string
+    - name: goal_object
+      dtype: string
+  splits:
+  - name: test
+    num_bytes: 10179370
+    num_examples: 100
+  download_size: 10179370
+  dataset_size: 10179370
+configs:
+- config_name: default
+  data_files:
+  - split: test
+    path: data/*
+task_categories:
+- visual-question-answering
+- image-to-text
+language:
+- en
+tags:
+- multimodal
+- chain-of-thought
+- reasoning
+- game
+- navigation
+size_categories:
+- n<1K
+---
 
-  - name: task
-    dtype: string
+# Game Reasoning Dataset
+
+A multimodal chain-of-thought reasoning dataset for game navigation and puzzle-solving tasks.
+
+## Dataset Description
+
+This dataset contains 100 visual reasoning examples from various game scenarios, featuring step-by-step reasoning with intermediate visual representations.
+
+### Supported Tasks
+
+- **Navigation**: Path planning in grid-based game environments
+- **Puzzle-solving**: Goal-oriented reasoning tasks
+
+## Dataset Structure
+
+Each example contains:
+
+| Field | Type | Description |
+|-------|------|-------------|
+| `pid` | string | Unique problem ID (`{scene}_{task}_{width}_{height}_{index}`) |
+| `question` | string | Task description |
+| `answer` | string | Expert path / solution |
+| `problem_image_0` | image | Initial game state visualization |
+| `resoning_thought_0` | string | First round reasoning text |
+| `reasoning_image_0` | image | Intermediate reasoning visualization |
+| `resoning_thought_1` | string | Second round reasoning text |
+| `full_text_only_thought` | string | Complete text-only reasoning chain |
+| `task` | string | Task identifier (`{scene}_{task}`) |
+| `meta_info` | dict | Metadata for verification |
+
+### Meta Info Structure
 
-  # ❗ DO NOT declare meta_info here
----
+```json
+{
+  "map": [[0, 1, 0], [0, 0, 1], [1, 0, 0]],  // 2D list representing grid layout
+  "icon_info": "Icon definitions and positions",
+  "task_type": "navigation",
+  "goal_object": "target object name"
+}