TEMPURA-Qwen2.5-VL-3B-s2

Model Card for Model ID

<!-- Provide a quick summary of what the model is/does. -->

This model card describes TEMPURA, a vision-language models to reason about causal event relationships and generate fine-grained, timestamped descriptions of untrimmed videos.

Model Details

Model Description

<!-- Provide a longer summary of what this model is. -->

TEMPURA enhances video temporal understanding by integrating causal reasoning with fine-grained temporal segmentation. More details can be found on the project page.

• Developed by: Jen-Hao Cheng, Vivian Wang, Huayu Wang, Huapeng Zhou, Yi-Hao Peng, Hou-I Liu, Hsiang-Wei Huang, Kuang-Ming Chen, Cheng-Yen Yang, Wenhao Chai, Yi-Ling Chen, Vibhav Vineet, Qin Cai, Jenq-Neng Hwang
• Model type: Video-Language Model
• Language(s) (NLP): English
• License: Qwen Research License (inherited from Qwen2.5-VL-3B-Instruct; non-commercial research use)
• Finetuned from model: Qwen/Qwen2.5-VL-3B-Instruct

Model Sources

<!-- Provide the basic links for the model. -->

• Repository: https://github.com/andy-cheng/TEMPURA
• Paper: TEMPURA: Temporal Event Masked Prediction and Understanding for Reasoning in Action
• Project Page: https://andy-cheng.github.io/TEMPURA/

Uses

<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->

Direct Use

<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->

The model can be used directly for temporal grounding and highlight detection in videos.

Downstream Use [optional]

<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->

The model can be fine-tuned for various applications requiring temporal video understanding, such as video summarization, event extraction, and question answering.

Out-of-Scope Use

<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->

The model may not perform well on videos with significantly different visual styles or languages compared to the training data.

Bias, Risks, and Limitations

<!-- This section is meant to convey both technical and sociotechnical limitations. -->

The model's performance is influenced by biases present in the VER dataset. Further analysis is needed to fully characterize these biases.

Recommendations

<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->

Users should be aware of potential biases in the model's outputs.

How to Get Started with the Model

Inference: Please check the inference example.

Training: Please check the model training script.

Training Details

Training Data

<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->

The model was trained on the VER dataset (https://huggingface.co/datasets/andaba/TEMPURA-VER).

Training Procedure

<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->

The training procedure involves masked event prediction and video event segmentation with temporal dense captioning. See the training scripts in the repository for details.

Training Hyperparameters

• Training regime: [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->

Speeds, Sizes, Times

<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->

[More Information Needed]

Evaluation

<!-- This section describes the evaluation protocols and provides the results. -->

Testing Data, Factors & Metrics

Testing Data

<!-- This should link to a Dataset Card if possible. -->

[More Information Needed]

Factors

<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->

[More Information Needed]

Metrics

<!-- These are the evaluation metrics being used, ideally with a description of why. -->

[More Information Needed]

Results

[More Information Needed]

Summary

Environmental Impact

<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->

Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).

• Hardware Type: [More Information Needed]
• Hours used: [More Information Needed]
• Cloud Provider: [More Information Needed]
• Compute Region: [More Information Needed]
• Carbon Emitted: [More Information Needed]

Technical Specifications [optional]

Model Architecture and Objective

[More Information Needed]

Compute Infrastructure

[More Information Needed]

Hardware

[More Information Needed]

Software

[More Information Needed]

Citation

<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->

BibTeX:

@article{tempura,
       title={TEMPURA: Temporal Event Masked Prediction and Understanding for Reasoning in Action}, 
       author={Jen-Hao Cheng and Vivian Wang and Huayu Wang and Huapeng Zhou and Yi-Hao Peng and Hou-I Liu
              and Hsiang-Wei Huang and Kuang-Ming Chen and Cheng-Yen Yang
              and Wenhao Chai and Yi-Ling Chen and Vibhav Vineet and Qin Cai and Jenq-Neng Hwang},
       journal={arXiv preprint arXiv:2505.01583},
       year={2025}
}

APA:

Cheng, J.-H., Wang, V., Wang, H., Zhou, H., Peng, Y.-H., Liu, H.-I., Huang, H.-W., Chen, K.-M., Yang, C.-Y., Chai, W., Chen, Y.-L., Vineet, V., Cai, Q., & Hwang, J.-N. (2025). TEMPURA: Temporal Event Masked Prediction and Understanding for Reasoning in Action. arXiv preprint arXiv:2505.01583.

Model Card Contact

Jen-Hao Cheng, [email protected]

License

These weights are a fine-tune of Qwen/Qwen2.5-VL-3B-Instruct and are released under the same Qwen Research License (a copy is included as LICENSE); non-commercial research use only, see the license for details.