Update README.md

README.md

@@ -1,59 +1,39 @@
 ---
+dataset_info:
+  features:
+  - name: id
+    dtype: string
+  - name: domain
+    dtype: string
+  - name: task_type
+    dtype: string
+  - name: prompt
+    dtype: string
+  - name: image
+    dtype: image
+  - name: reference_frames
+    sequence: image
+  - name: reference_text
+    sequence: string
+  - name: protocol
+    dtype: string
+configs:
+- config_name: default
+  data_files:
+  - split: train
+    path: dataset.parquet
 license: mit
 language:
 - en
-size_categories:
-- n<1K
 ---
-# VIPER Benchmark
 
-## Dataset Description
+# Beyond the Last Frame: Process-aware Evaluation for Generative Video Reasoning
 
-VIPER (VIdeo Process Evaluation for Reasoning tasks) is a comprehensive benchmark for Generative Video Reasoning (GVR) tasks. 
+## 👀 About VIPER
 
-### Dataset Summary
-
-This dataset contains **309 samples** spanning **6 distinct reasoning domains**:
-
-- **Physics**: Understanding of physical laws and dynamics (32 samples)
-- **Planning**: Spatial navigation and object manipulation (44 samples)
-- **Spatial**: 3D spatial reasoning and transformations (60 samples)
-- **Structural**: Pattern recognition and constraint satisfaction (70 samples)
-- **Symbolic**: Knowledge reasoning and mathematical problem-solving (60 samples)
-- **Temporal**: Temporal dynamics and object tracking (43 samples)
-
-## Dataset Structure
-
-### Data Instances
-
-Each instance contains:
-
-```json
-{
-  "domain": "Spatial",
-  "id": "block_rotate_0",
-  "image": "Spatial/images/block_rotate_0.jpg",
-  "reference": {
-    "frames": ["Spatial/references/block_rotate_0_0.jpg"],
-    "text": []
-  },
-  "prompt": "A structure composed of 9 cubes viewed from...",
-  "protocol": ["Maintain the block structure..."],
-  "task_type": "block_rotate"
-}
-```
-
-### Data Fields
-
-- `domain`: The reasoning domain (Physics, Planning, Spatial, Structural, Symbolic, Temporal)
-- `id`: Unique identifier for the sample
-- `image`: Path to the input image
-- `reference`: Reference materials for evaluation
-  - `frames`: Reference image frames
-  - `text`: Reference text descriptions
-- `prompt`: Text prompt describing the task or desired output
-- `protocol`: Evaluation criteria or constraints
-- `task_type`: Specific task category within the domain
+- Overview: Process-aware evaluation for Generative Video Reasoning tasks.
+- Statistics: 309 carefully curated samples spanning 6 distinct domains (i.e., temporal, structural, symbolic, spatial, physics and planning reasoning).
+- New Metric: Process-outcome Consistency (POC@$r$). POC@$r$ evaluate video correctness at both process- and outcome-level, with multiple frames uniformly sampled from the whole video at rate $r$, instead of the last frame only.
 
 
 ## Dataset Statistics
@@ -69,37 +49,9 @@ Each instance contains:
 | Symbolic | 60 | knowledge, math, multimodal |
 | Temporal | 43 | obj_move, zoom |
 
-### Task Distribution by Domain
-
-**Physics** (32 samples):
-- experiment: 18 samples
-- game: 14 samples
+## 📦 Dataset Usage
 
-**Planning** (44 samples):
-- navigation: 25 samples
-- obj_manipulation: 19 samples
-
-**Spatial** (60 samples):
-- block_rotate: 25 samples
-- dice: 20 samples
-- image_restore: 15 samples
-
-**Structural** (70 samples):
-- chess: 20 samples
-- maze: 20 samples
-- sudoku: 20 samples
-- ttt: 10 samples
-
-**Symbolic** (60 samples):
-- knowledge: 20 samples
-- math: 20 samples
-- multimodal: 20 samples
-
-**Temporal** (43 samples):
-- obj_move: 25 samples
-- zoom: 18 samples
-
-## Usage
+### Download
 
 ```python
 from datasets import load_dataset
@@ -109,7 +61,22 @@ dataset = load_dataset("Monosail/VIPER")
 
 ```
 
-## Citation Information
+### Data Fields
+
+- `id`: Unique identifier for the sample
+- `domain`: The reasoning domain (Physics, Planning, Spatial, Structural, Symbolic, Temporal)
+- `task_type`: Specific task category within the domain
+- `prompt`: Text prompt describing the task
+- `image`: The input image
+- `reference_frames`: Ground-truth image frames
+- `reference_texts`: Ground-truth text descriptions
+- `protocol`: Process-level task constraints
+
+
+
+## 📝 Citation
+
+If you find our benchmark useful, please consider citing us:
 
 ```bibtex
 @article{li2026viper,