|
|
--- |
|
|
task_categories: |
|
|
- image-text-to-text |
|
|
- video-text-to-text |
|
|
- object-detection |
|
|
- image-segmentation |
|
|
language: |
|
|
- en |
|
|
--- |
|
|
|
|
|
This repository contains the evaluation data presented in: [OneThinker: All-in-one Reasoning Model for Image and Video](https://arxiv.org/abs/2512.03043) |
|
|
|
|
|
Project Page: https://github.com/tulerfeng/OneThinker |
|
|
Code: https://github.com/tulerfeng/OneThinker |
|
|
|
|
|
## About OneThinker |
|
|
|
|
|
<div align="center"> |
|
|
<img src="https://github.com/tulerfeng/OneThinker/blob/main/assets/teaser.png?raw=true" alt="OneThinker Teaser" width="95%"> |
|
|
</div> |
|
|
|
|
|
We introduce **OneThinker**, an all-in-one multimodal reasoning generalist that is **capable of thinking across a wide range of fundamental visual tasks within a single model**. |
|
|
|
|
|
We construct the large-scale **OneThinker-600k** multi-task training corpus and build **OneThinker-SFT-340k** with high-quality CoT annotations for cold-start SFT. Moreover, we propose **EMA-GRPO**, a new RL method that **balances heterogeneous reward signals across diverse visual tasks**, via simply tracking task-wise moving averages of reward std. |
|
|
|
|
|
OneThinker demonstrates **strong performance on 31 benchmarks across 10 fundamental vision tasks**, while showing cross-task knowledge transfer and promising zero-shot generalization toward a **unified multimodal reasoning generalist**. |
|
|
|
|
|
All code, models, and data are fully released. |
|
|
|
|
|
## Dataset |
|
|
|
|
|
Our dataset covers both image and video modalities and spans a series of fundamental visual reasoning tasks, including rule-based QA, open-ended QA, captioning, spatial grounding, temporal grounding, spatio-temporal grounding, tracking, and segmentation |
|
|
|
|
|
<div align="center"> |
|
|
<img src="https://github.com/tulerfeng/OneThinker/blob/main/assets/dataset.png?raw=true" alt="OneThinker Dataset Overview" width="90%"> |
|
|
</div> |
|
|
|
|
|
To enable effective SFT initialization for reasoning, we leverage a strong proprietary model, Seed1.5-VL to produce CoT annotations. |
|
|
|
|
|
The `onethinker_rl_train.json` file is for RL training while `onethinker_sft_image.json` and `onethinker_sft_video.json` is for SFT cold start. The json files end with `_unsampled` are unsampled full set. |
|
|
|
|
|
## Sample Usage |
|
|
|
|
|
For inference on a single example, you may refer to: |
|
|
|
|
|
```bash |
|
|
python ./Evaluation/inference_single/inference.py |
|
|
``` |
