Update dataset card

README.md

@@ -50,11 +50,11 @@ dataset_info:
 
 # K600-MM
 
-K600-MM is a multimodal video dataset used to pretrain [A2A-Video](https://a2a-video.epfl.ch/), an any-to-any multimodal model for the video domain. It's built on top of [Kinetics-600](https://www.deepmind.com/open-source/kinetics) (RGB video + class-category annotations) with 10 additional modalities obtained via pseudo-labeling. Refer to [A2A-Video's README_DATA.md](https://github.com/EPFL-VILAB/A2A-Video/blob/main/README_DATA.md) for additional details on dataset construction.
+K600-MM is a multimodal video dataset used to pretrain [Video-4M](https://video-4m.epfl.ch/), an any-to-any multimodal model for the video domain. It's built on top of [Kinetics-600](https://www.deepmind.com/open-source/kinetics) (RGB video + class-category annotations) with 10 additional modalities obtained via pseudo-labeling. Refer to [Video-4M's README_DATA.md](https://github.com/EPFL-VILAB/Video-4M/blob/main/README_DATA.md) for additional details on dataset construction.
 
 In total, there are ~392K training and ~30K validation/test video clips with 12 aligned modalities per clip.
 
-**The provided train/test data is tokenized**: every modality except textual ones is already tokenized using its corresponding tokenizer. Samples are sharded as WebDataset `.tar` files per modality (see [Directory structure](#directory-structure)). See [README_DATA.md](https://github.com/EPFL-VILAB/A2A-Video/blob/main/README_DATA.md) for how to obtain the raw videos and set up the full dataset.
+**The provided train/test data is tokenized**: every modality except textual ones is already tokenized using its corresponding tokenizer. Samples are sharded as WebDataset `.tar` files per modality (see [Directory structure](#directory-structure)). See [README_DATA.md](https://github.com/EPFL-VILAB/Video-4M/blob/main/README_DATA.md) for how to obtain the raw videos and set up the full dataset.
 
 Samples are aligned by filename across separate per-modality folders. Browse the **`samples`** tab above for a small preview from the test split.
 
@@ -90,7 +90,7 @@ snapshot_download(
 
 ## Visualizing the dataset
 
-Each modality is tokenized with its own dedicated tokenizer, hosted at [🤗 EPFL-VILAB/A2A-Video-tokenizers](https://huggingface.co/EPFL-VILAB/A2A-Video-tokenizers). To decode and visualize samples yourself, run `dataset_visualization/visualize_k600mm_samples.py` from the [A2A-Video repo](https://github.com/EPFL-VILAB/A2A-Video):
+Each modality is tokenized with its own dedicated tokenizer, hosted at [🤗 EPFL-VILAB/Video-4M-tokenizers](https://huggingface.co/EPFL-VILAB/Video-4M-tokenizers). To decode and visualize samples yourself, run `dataset_visualization/visualize_k600mm_samples.py` from the [Video-4M repo](https://github.com/EPFL-VILAB/Video-4M):
 
 ```bash
 # tokenizer checkpoints are loaded automatically from the hub
@@ -101,20 +101,20 @@ python dataset_visualization/visualize_k600mm_samples.py \
     --output_dir ./dataset_visualization_output
 ```
 
-## Using K600-MM to pretrain A2A-Video
+## Using K600-MM to pretrain Video-4M
 
-Refer to the [A2A-Video GitHub repository](https://github.com/EPFL-VILAB/A2A-Video) for details on using K600-MM for pretaining A2A-Video models.
+Refer to the [Video-4M GitHub repository](https://github.com/EPFL-VILAB/Video-4M) for details on using K600-MM for pretraining Video-4M models.
 
 ## License
 
-Released under Apache 2.0. Kinetics-600 (the underlying source video) and the pseudo-labeling models used to derive the other modalities carry their own terms. 
+Released under Apache 2.0. Kinetics-600 (the underlying source video) and the pseudo-labeling models used to derive the other modalities carry their own terms.
 
 ## Citation
 
 ```
-@article{a2avideo2026,
-    title={{A2A-Video: Modeling the World Across Time and Modalities}},
-    author={Khattak, Muhammad Uzair and Kim, Won Jun and Abbassi, Reza and Havolli, Albias and Murphy, Michael and Zadeh, Amir and Li, Chuan and Kar, O\u{g}uzhan Fatih and Naeem, Muhammad Ferjad and Atanov, Andrei and Bachmann, Roman and Tombari, Federico and Zamir, Amir},
+@article{video4m2026,
+    title={{Any-to-Any Video Modeling: Modeling the World by Traversing Time and Modalities}},
+    author={Khattak, Muhammad Uzair and Kim, Won Jun and Abbassi, Reza and Havolli, Albias and Murphy, Michael and Zadeh, Amir and Li, Chuan and Naeem, Muhammad Ferjad and Kar, O\u{g}uzhan Fatih and Bachmann, Roman and Atanov, Andrei and Tombari, Federico and Zamir, Amir},
     journal={arXiv preprint},
     year={2026},
 }