README.md

metadata

license: cc-by-nc-sa-4.0
configs:
  - config_name: video_perspective
    data_files: video_perspective.json
  - config_name: question_perspective
    data_files: question_perspective.json
  - config_name: train
    data_files: train.json
task_categories:
  - video-text-to-text

A Comprehensive Benchmark for Fine-Grained Video Motion Understanding

---

🔥 News

• 2025.09.18 🎉 FAVOR-Bench has been accepted by NeurIPS 2025 Datasets and Benchmarks Track!
• 2025.03.19 🌟 We released Favor-Bench, a new benchmark for fine-grained video motion understanding that spans both ego-centric and third-person perspectives with comprehensive evaluation including both close-ended QA tasks and open-ended descriptive tasks!

Introduction

Multimodal Large Language Models (MLLMs) have shown impressive video content understanding capabilities but struggle with fine-grained motion comprehension. To comprehensively assess the motion understanding ability of existing MLLMs, we introduce FAVOR-Bench, which comprises 1,776 videos from both ego-centric and third-person perspectives and enables assessment through both close-ended and open-ended tasks. For close-ended evaluation, we carefully design 8,184 multiple-choice question-answer pairs spanning six distinct sub-tasks. For open-ended evaluation, we employ the GPT-assisted evaluation and develop a novel cost-efficient LLM-free assessment method, where the latter can enhance benchmarking interpretability and accessibility. Comprehensive experiments with 21 state-of-the-art MLLMs reveal significant limitations in their ability to comprehend and describe detailed temporal dynamics in video motions. To alleviate this limitation, we further build FAVOR-Train, a dataset of 17,152 videos with fine-grained motion annotations. Finetuning Qwen2.5-VL on FAVOR-Train yields consistent improvements on motion-related tasks across TVBench, MotionBench and our FAVOR-Bench. Our assessment results demonstrate that the proposed FAVOR-Bench and FAVOR-Train provide valuable tools for the community to develop more powerful video understanding models.

Evaluation Tasks

Dataset

License

Our dataset is under the CC-BY-NC-SA-4.0 license.

If you need to access and use our dataset, you must understand and agree: This dataset is for research purposes only and cannot be used for any commercial or other purposes. The user assumes all effects arising from any other use and dissemination.

We do not own the copyright of any raw video files. Currently, we provide video access to researchers under the condition of acknowledging the above license. For the video data used, we respect and acknowledge any copyrights of the video authors. Therefore, for the TV series and animations used in the dataset, we have applied several preprocessing steps to minimize any potential impact on the original copyrights. These include reducing video resolution, segmenting videos into short clips (less than 10 seconds), and applying dimension adjustments.

If there is any infringement in FAVOR-Bench, please contact zhangl22@m.fudan.edu.cn or directly raise an issue. If necessary, we will replace the contested videos with sparsely sampled frame sets at adjusted resolutions. In cases where even frame retention is not permitted, we will maintain the annotation files while replacing the video content with meta-information or seeking alternative sources that are more reliable and risk-free.

FAVOR-Bench Videos

We provide all self-collected video clips from TV series and animations in this space.

For publically available videos, you could download them from the original address:

1. Charades: https://prior.allenai.org/projects/charades
2. EgoTaskQA: https://sites.google.com/view/egotaskqa

FAVOR-Train Videos

For videos originated from Koala36M, we provide their Youtube links and start&end time. You could download them with tools like yt-dlp.

For publically available videos, you could download them from the original address:

1. Charades-ego: https://prior.allenai.org/projects/charades-ego
2. EgoTaskQA: https://sites.google.com/view/egotaskqa
3. EgoExoLearn: https://huggingface.co/datasets/hyf015/EgoExoLearn
4. EgoExo4D: https://ego-exo4d-data.org/

For EgoExoLearn and EgoExo4D, you can crop the original videos according the start&end time provided in the JSON file by yourself.

JSON Files

For FAVOR-Bench, we provide both question-perspective and video-perspective dicts.

In the video-perspective file, each entry represents one video and we provide caption, camera motion, subject attributes, motion list, chronological motion list and all questions (question, options, correct answer, task type).

In question perspective, each entry represents a single question, including question, options, correct answer, task type, and the corresponding video name.

📈 Results

• Model Comparision:

• Benchmark Comparison:

• Benchmark Statistics:

Data statistics of FAVOR-Bench. Left: Task type distribution across close-ended and open-ended evaluation in FAVOR-Bench. Middle: Distribution of motion numbers (motion sequence length) per video. Right: The word cloud statistics of motion vocabularies in FAVOR-Bench.

More data statistics of FAVOR-Bench. Left: Index distribution of correct answers for the close-ended tasks. For example, "(1)" indicates that the correct option is ranked first. Middle: Video duration distribution of FAVOR-Bench. Right: Question number distribution for videos of FAVOR-Bench.

Citation

If you find our work helpful for your research, please consider citing our work.

@article{tu2026favor,
  title={Favor-bench: A comprehensive benchmark for fine-grained video motion understanding},
  author={Tu, Chongjun and Zhang, Lin and Ye, Peng and Zeng, Xianfang and Cheng, Wei and Yu, Gang and Chen, Tao and others},
  journal={Advances in Neural Information Processing Systems},
  volume={38},
  year={2026}
}