README.md

---
license: cc-by-nc-4.0
configs:
- config_name: mcq
  data_files:
  - split: test
    path: sember_mcq.jsonl
- config_name: grounding
  data_files:
  - split: test
    path: sember_grounding.jsonl
extra_gated_prompt: >-
  S-EMBER is released for non-commercial research purposes only. By requesting
  access you agree to use this dataset solely for non-commercial purposes, in
  accordance with the Creative Commons Attribution-NonCommercial 4.0
  International (CC BY-NC 4.0) license.
extra_gated_fields:
  Full name: text
  Affiliation: text
  Intended use: text
  I agree to use this dataset for non-commercial purposes only: checkbox
  I agree to the CC BY-NC license terms: checkbox
---

# S-EMBER

Episodic-memory video QA benchmark (face-blurred, audio-removed).

## License & usage

This dataset is licensed under
[CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/) and is provided
for **non-commercial research use only**. Access is gated: you must accept the
non-commercial terms above before downloading.

## Contents

- `sember_mcq.jsonl` — multiple-choice evaluation split.
- `sember_grounding.jsonl` — answer-generation and temporal-grounding evaluation split.
- `videos/` — face-blurred, audio-removed `.mp4` clips (720p, 12fps).

Each JSONL row has a `video_id` field that maps to `videos/<video_id>.mp4`.

## Loading with Hugging Face Datasets

```python
from datasets import load_dataset

mcq = load_dataset("xidwang/S-EMBER", "mcq", split="test")
grounding = load_dataset("xidwang/S-EMBER", "grounding", split="test")
```