README.md

metadata

task_categories:
  - image-text-to-text
  - video-text-to-text
  - audio-text-to-text

ThinkOmni Evaluation Dataset

This repository contains the evaluation datasets for ThinkOmni, a training-free framework that lifts textual reasoning to omni-modal scenarios via guidance decoding.

ThinkOmni enhances omni-modal large language models (OLLMs) with the reasoning capabilities of large reasoning models (LRMs) at decoding time, adaptively balancing perception and reasoning signals.

• Paper: ThinkOmni: Lifting Textual Reasoning to Omni-modal Scenarios via Guidance Decoding
• Repository: https://github.com/1ranGuan/thinkomni

Dataset Contents

This collection includes the following evaluation sets used in the paper:

• MMAU
• OmniBench
• Daily-Omni

Citation

If you find this work or these datasets useful, please cite:

@inproceedings{guan2026thinkomni,
  title={ThinkOmni: Lifting Textual Reasoning to Omni-modal Scenarios via Guidance Decoding},
  author={Guan, Yiran and Tu, Sifan and Liang, Dingkang and Zhu, Linghao and Ju, Jianzhong and Luo, Zhenbo and Luan, Jian and Liu, Yuliang and Bai, Xiang},
  booktitle={International Conference on Learning Representations (ICLR)},
  year={2026}
}