anonupload/EgoGazeVQA-91-nips25DB

EgoGazeVQA-91 • NeurIPS 2025 Datasets & Benchmarks submission

In the Eye of MLLM: Benchmarking Egocentric Video Intent Understanding with Gaze-Guided Prompting

---

1 Folder layout

EgoGazeVQA-91-nips25DB/         
├── qa\_pairs/             # VQA supervision
│   ├── causal\_ego4d.{csv,json}
│   ├── spatial\_ego4d.{csv,json}
│   ├── temporal\_ego4d.{csv,json}
│   └── ...
├── keyframe_tar/            
│   ├── ego4d.tar.gz
│   ├── egoexo.tar.gz
│   └── egtea.tar.gz
└── keyframes/
    ├── ego4d
    ├── egoexo
    └── egtea
          ├── video\_id
          └── ...
  

---

2 Quick start — load key-frames in Python

from datasets import load_dataset, DownloadConfig


ds = load_dataset(
    "anonupload/EgoGazeVQA-91-nips25DB",
    data_dir="keyframes/ego4d",           # only key-frames
    split="train",
    )
)

print("Number of key-frames:", len(ds))
print(ds[0]["image"].shape)            # PIL Image

---

3 Licence

All files are released under CC-BY-4.0. You are free to use, share, and adapt the data — please cite the paper when you do.

---

4 Ethics & privacy

Videos were captured in controlled environments with written consent from all recorded participants. Faces and private content have been manually reviewed and blurred where necessary. Please report any remaining privacy issue via the repo’s “Report this dataset” button.

---