LongTVQA Dataset Files

This repository contains the LongTVQA dataset exports as JSON-formatted files. Despite the .jsonl suffix, the QA splits are stored as a JSON array, and the subtitle files are JSON objects.

Data files

• LongTVQA_train.jsonl — training split QA list.
• LongTVQA_val.jsonl — validation split QA list.
• LongTVQA_subtitles_clip_level.jsonl — clip-level subtitle text indexed by occur_clip (e.g. castle_s01e01_seg02_clip_00).
• LongTVQA_subtitles_episode_level.jsonl — episode-level subtitle text indexed by episode_name (e.g. castle_s01e01).

Subtitle file examples

Clip-level subtitles map clip ids to a single subtitle string:

"castle_s01e01_seg02_clip_00": "(Kath...)"

Episode-level subtitles map episode ids to a string that concatenates clips with segment markers:

"castle_s01e01": "<seg02_clip_00>xxx</seg02_clip_00>xxx"

QA item schema

Each entry in the train/val arrays is a QA item with keys such as:

• qid: integer question id.
• q: question text.
• a0-a4: answer options.
• answer: correct option key ("a0"-"a4").
• ts: [start, end] float list of the temporal span.
• episode_name: episode identifier (e.g. grey_s03e20).
• occur_clip: clip identifier (e.g. grey_s03e20_seg02_clip_14).
• show_name: show title.

📝 Citation

If you find our work helpful, please cite:

@misc{liu2025longvideoagentmultiagentreasoninglong,
      title={LongVideoAgent: Multi-Agent Reasoning with Long Videos}, 
      author={Runtao Liu and Ziyi Liu and Jiaqi Tang and Yue Ma and Renjie Pi and Jipeng Zhang and Qifeng Chen},
      year={2025},
      eprint={2512.20618},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={[https://arxiv.org/abs/2512.20618](https://arxiv.org/abs/2512.20618)}, 
}