Skip to content

Visual-AI/iVGR

Repository files navigation

iVGR: Internalizing Visually Grounded Reasoning for MLLMs with Reinforcement Learning

Conference Paper Project HuggingFace

Method

iVGR method overview

Qualitative Results

Grounded CoT vs Textual CoT within iVGR

Environment Setup
# TODO: add installation instructions
Data Preparation
# TODO: describe how to download / preprocess the training and evaluation data
Training
bash examples/train_iVGR_qwen2_5_VL.sh
Evaluation
# TODO: add evaluation commands
Acknowledgements

TODO: list the projects / codebases this work builds on (e.g., verl, Qwen2.5-VL, TreeVGR, ...).

BibTeX
@inproceedings{zhang2026ivgr,
  title     = {iVGR: Internalizing Visually Grounded Reasoning for MLLMs with Reinforcement Learning},
  author    = {Zhang, Chang-Bin and Zhong, Yujie and Zhang, Qiang and Han, Kai},
  booktitle = {International Conference on Machine Learning (ICML)},
  year      = {2026}
}

About

[ICML 2026] iVGR: Internalizing Visually Grounded Reasoning for MLLMs with Reinforcement Learning

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors