diff --git a/README.md b/README.md
index a2543cd..213e102 100644
--- a/README.md
+++ b/README.md
@@ -11,6 +11,7 @@
 
 <!-- [[`📕 Paper`](https://arxiv.org/abs/2307.04657)] -->
 [[`🏠 Project Homepage`](https://sites.google.com/view/safe-sora)]
+[[`📕 Paper`](https://arxiv.org/abs/2406.14477)]
 [[`🤗 SafeSora Datasets`](https://huggingface.co/datasets/PKU-Alignment/SafeSora)]
 [[`🤗 SafeSora Label`](https://huggingface.co/datasets/PKU-Alignment/SafeSora-Label)]
 [[`🤗 SafeSora Evaluation`](https://huggingface.co/datasets/PKU-Alignment/SafeSora-Eval)]
@@ -140,11 +141,13 @@ eval_data = PromptDataset.load("path/to/config", video_dir="path/to/video_dir")
 If you find the SafeSora dataset family useful in your research, please cite the following paper:
 
 ```bibtex
-@article{SafeSora2024,
-  title   = {SafeSora: Towards Safety Alignment of Text2Video Generation via a Human Preference Dataset},
-  author  = {Josef Dai and Tianle Chen and Xuyao Wang and Ziran Yang and Taiye Chen and Jiaming Ji and Yaodong Yang},
-  url     = {https://github.com/calico-1226/safe-sora},
-  year    = {2024}
+@misc{dai2024safesora,
+      title={SafeSora: Towards Safety Alignment of Text2Video Generation via a Human Preference Dataset},
+      author={Josef Dai and Tianle Chen and Xuyao Wang and Ziran Yang and Taiye Chen and Jiaming Ji and Yaodong Yang},
+      year={2024},
+      eprint={2406.14477},
+      archivePrefix={arXiv},
+      primaryClass={id='cs.CV' full_name='Computer Vision and Pattern Recognition' is_active=True alt_name=None in_archive='cs' is_general=False description='Covers image processing, computer vision, pattern recognition, and scene understanding. Roughly includes material in ACM Subject Classes I.2.10, I.4, and I.5.'}
 }
 ```
 
diff --git a/pyproject.toml b/pyproject.toml
index a434acc..d9647a1 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -38,6 +38,7 @@ dependencies = [
     "transformers",
     "datasets",
     "tokenizers",
+    "av",
 ]
 dynamic = ["version"]
 
diff --git a/safe_sora/datasets/base.py b/safe_sora/datasets/base.py
index 288ec46..d929d5d 100644
--- a/safe_sora/datasets/base.py
+++ b/safe_sora/datasets/base.py
@@ -40,9 +40,7 @@
 
 def is_complete(data_dict: dict) -> bool:
     """Check if a dictionary is complete, i.e., all values are not None."""
-    for key, value in data_dict.items():
-        if key == 'info':
-            continue
+    for _, value in data_dict.items():
         if isinstance(value, dict) and not is_complete(value):
             return False
         if value is None:
@@ -153,7 +151,6 @@ class VideoSample(TypedDict):
     is_safe: NotRequired[bool]
     video_labels: NotRequired[HarmLabel]
     generated_from: NotRequired[str]
-    info: NotRequired[dict]
 
 
 def format_video_sample_from_dict(data: dict, contain_labels: bool = False) -> VideoSample:
@@ -183,7 +180,6 @@ def format_video_sample_from_dict(data: dict, contain_labels: bool = False) -> V
             is_safe=data.get('is_safe'),
             video_labels=video_labels,
             generated_from=data.get('generated_from'),
-            info=data.get('info', {}),
         )
 
     return VideoSample(
@@ -195,7 +191,6 @@ def format_video_sample_from_dict(data: dict, contain_labels: bool = False) -> V
         video_path=data.get('video_path'),
         is_safe=data.get('is_safe'),
         generated_from=data.get('generated_from'),
-        info=data.get('info', {}),
     )
 
 
@@ -236,7 +231,6 @@ class VideoPairSample(TypedDict):
     helpfulness: NotRequired[Literal['video_0', 'video_1']]
     harmlessness: NotRequired[Literal['video_0', 'video_1']]
     sub_preferences: NotRequired[SubPreference]
-    info: NotRequired[dict[str, str]]
 
 
 def format_video_pair_sample_from_dict(data: dict) -> VideoPairSample:
@@ -270,7 +264,6 @@ def format_video_pair_sample_from_dict(data: dict) -> VideoPairSample:
         helpfulness=data.get('helpfulness'),
         harmlessness=data.get('harmlessness'),
         sub_preferences=sub_preferences,
-        info=data.get('info', {}),
     )