LongVideoAgent Qwen2.5-7B

This repository hosts the released LLM checkpoint for LongVideoAgent, a multi-agent framework for long-video question answering. This model is a Qwen2.5-7B-based checkpoint used in the LongVideoAgent project.

Project links:

Project page: https://longvideoagent.github.io/
Code: https://github.com/longvideoagent/LongVideoAgent
Paper: https://arxiv.org/abs/2512.20618

Overview

LongVideoAgent decomposes long-video reasoning into multiple roles:

MasterAgent for planning and answer generation
GroundingAgent for subtitle-based temporal grounding
VisionAgent for local visual evidence extraction

This checkpoint is intended for use with the official LongVideoAgent codebase and evaluation pipeline.

LongTVQA+ Performance

Compared with the Qwen2.5-7B-Instruct baseline, this checkpoint improves LongTVQA+ accuracy by 6.66 percentage points.

Model	LongTVQA+ Acc	Delta
Qwen2.5-7B-Instruct	57.33%	-
LongVideoAgent Qwen2.5-7B	64.00%	+6.66%

Intended Use

Use this model for:

Research on long-video question answering
Reproducing LongVideoAgent experiments
Studying agentic reasoning over long videos

This checkpoint is not a general-purpose video model by itself. For inference and evaluation, please use the official repository:

https://github.com/longvideoagent/LongVideoAgent

Usage

Please follow the setup and inference instructions in the official repository and project documentation:

If you use this checkpoint in your work, please also cite the LongVideoAgent paper below.

Citation

@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},
}