nvidia/VideoITG-8B

1---2license: other3license_name: nvlicense4license_link: LICENSE5pipeline_tag: image-text-to-text6library_name: transformers7base_model:8- Qwen/Qwen2-7B-Instruct9- google/siglip-so400m-patch14-38410base_model_relation: merge11language:12- multilingual13tags:14- VideoITG15- Eagle16- VLM17---18 19# VideoITG-8B20[\[🌐Homepage\]](https://nvlabs.github.io/VideoITG/)   [\[💻GitHub\]](https://github.com/NVlabs/VideoITG)   [\[📜Tech Report\]](https://arxiv.org/abs/2507.13353)21[\[🤗VideoITG-40K\]](https://huggingface.co/datasets/NVEagle/VideoITG-40K)22 23## Introduction24 25VideoITG-8B is a multimodal video understanding model trained with instructed temporal grounding, equipped with the ability to enhance Video Large Language Models through intelligent frame selection. The model tackles the complexities of real-world video scenarios by aligning frame sampling with user instructions. Please check our paper for more details.26 27## Model Details28 29- **Model name**: VideoITG-8B30- **Architecture**: Customized Eagle-8B base model, fine-tuned with Instructed Temporal Grounding31- **Model type**: Multimodal Large Language Model with Video Understanding32- **Languages**: English (primary), multilingual (partially)33 34## Model Performance35 36| Model               | Base Model        | Frames | LongVideoBench | MLVU | VideoMME | CG-Bench |37|---------------------|-------------------|--------|----------------|------|----------|----------|38| VideoITG-7B         | InternVL2.5-8B    | 32     | 61.9 (+2.9%)   | 75.0 (+7.8%) | 67.3 (+4.0%) | 46.7 (+7.0%) |39| VideoITG-7B         | InternVL2.5-26B   | 32     | 63.0 (+1.0%)   | 78.9 (+6.1%) | 69.9 (+2.5%) | 48.7 (+6.0%) |40| VideoITG-7B         | LLaVA-Video-7B    | 32     | 61.6 (+3.6%)   | 74.6 (+8.6%) | 66.1 (+3.0%) | 42.8 (+9.0%) |41| VideoITG-7B         | LLaVA-Video-7B    | 64     | 60.9 (+7.4%)   | 76.3 (+7.6%) | 66.4 (+1.9%) | 42.9 (+8.1%) |42 43## Key Features44 45- **Instructed Temporal Grounding**: Intelligently selects video frames based on user instructions46- **Plug-and-Play**: Seamlessly integrates with existing video language models47- **Superior Temporal Understanding**: Excels in tasks requiring precise temporal grounding48 49 50## License51 52- Code: [Apache 2.0 License](https://www.apache.org/licenses/LICENSE-2.0)53- Model: [NVIDIA License](LICENSE) - Research preview for non-commercial use only54 55## Citation56 57If you find this project useful, please cite our work:58 59```bibtex60@article{wang2025videoitg,61  title     = {VideoITG: Multimodal Video Understanding with Instructed Temporal Grounding},62  author    = {Shihao Wang and Guo Chen and De-An Huang and Zhiqi Li and Minghan Li and Guilin Liu and Jose M. Alvarez and Lei Zhang and Zhiding Yu},63  journal   = {arXiv preprint arXiv:2507.13353},64  year      = {2025}65}66```67 68## Acknowledgement69 70- [Eagle](https://github.com/NVlabs/EAGLE): The codebase we built upon71- [LMMs-Eval](https://github.com/EvolvingLMMs-Lab/lmms-eval): Many thanks to the LMMs-Lab for the easy-to-use evaluation tools72- [LLaVA-OneVision](https://huggingface.co/datasets/lmms-lab/LLaVA-OneVision-Data) and [LLaVA-Video](https://huggingface.co/datasets/lmms-lab/LLaVA-Video-178K): We train our models with data from these great open-source projects73