Traceable Evidence Enhanced Visual Grounded Reasoning: Evaluation and Methodology

TL; DR: We propose TreeBench, the first benchmark specially designed for evaluating "thinking with images" capabilities with traceable visual evidence, and TreeVGR, the current state-of-the-art open-source visual grounded reasoning models.

Abstract. Models like OpenAI-o3 pioneer visual grounded reasoning by dynamically referencing visual regions,

just like human "thinking with images". However, no benchmark exists to evaluate these capabilities

holistically. To bridge this gap, we propose TreeBench (Traceable Evidence Evaluation Benchmark),

a diagnostic benchmark built on three principles: (1) focused visual perception of subtle targets in

complex scenes, (2) traceable evidence via bounding box evaluation, and (3) second-order reasoning

to test object interactions and spatial hierarchies beyond simple object localization. Prioritizing

images with dense objects, we initially sample 1K high-quality images from SA-1B, and incorporate

eight LMM experts to manually annotate questions, candidate options, and answers for each

image. After three stages of quality control, TreeBench consists of 405 challenging visual question-

answering pairs, even the most advanced models struggle with this benchmark, where none of

them reach 60% accuracy, e.g., OpenAI-o3 scores only 54.87. Furthermore, we introduce TreeVGR

(Traceable Evidence Enhanced Visual Grounded Reasoning), a training paradigm to supervise

localization and reasoning jointly with reinforcement learning, enabling accurate localizations and

explainable reasoning pathways. Initialized from Qwen2.5-VL-7B, it improves V* Bench (+16.8),

MME-RealWorld (+12.6), and TreeBench (+13.4), proving traceability is key to advancing visual

grounded reasoning.

Release

• [2026/01/26] 🔥🔥🔥 TreeBench and TreeVGR have been accepted by ICLR 2026! 🔥🔥🔥
• [2025/08/12] 🔥 TreeBench has been utilized by GLM-4.5V!
• [2025/07/12] 🔥🔥🔥 TreeBench and TreeVGR have been supported by VLMEvalKit! 🔥🔥🔥
• [2025/07/11] 🔥 TreeBench and TreeVGR have been released. Check out the paper for details.

Installation

pip3 install -r requirements.txt
pip3 install flash-attn --no-build-isolation -v

Usage

This repo provides a simple local inference demo of our TreeVGR on TreeBench. First, clone this repo,

git clone https://github.com/Haochen-Wang409/TreeVGR cd TreeVGR

python3 inference_treebench.py

This should give:

Perception/Attributes 18/29=62.07 Perception/Material 7/13=53.85 Perception/Physical State 19/23=82.61 Perception/Object Retrieval 10/16=62.5 Perception/OCR 42/68=61.76 Reasoning/Perspective Transform 19/85=22.35 Reasoning/Ordering 20/57=35.09 Reasoning/Contact and Occlusion 25/41=60.98 Reasoning/Spatial Containment 20/29=68.97 Reasoning/Comparison 20/44=45.45 ==> Overall 200/405=49.38 ==> Mean IoU: 43.3

Benchmark

• TreeBench: https://huggingface.co/datasets/HaochenWang/TreeBench

• TreeVGR-7B: https://huggingface.co/HaochenWang/TreeVGR-7B
• TreeVGR-7B-CI: https://huggingface.co/HaochenWang/TreeVGR-7B-CI

• TreeVGR-RL-37K: https://huggingface.co/datasets/HaochenWang/TreeVGR-RL-37K
• TreeVGR-SFT-35K: https://huggingface.co/datasets/HaochenWang/TreeVGR-SFT-35K

Citation

If you find Ross useful for your research and applications, please cite using this BibTeX:

@article{wang2025traceable,   title={Traceable Evidence Enhanced Visual Grounded Reasoning: Evaluation and Methodology},   author={Haochen Wang and Xiangtai Li and Zilong Huang and Anran Wang and Jiacong Wang and Tao Zhang and Jiani Zheng and Sule Bai and Zijian Kang and Jiashi Feng and Zhuochen Wang and Zhaoxiang Zhang},   journal={arXiv preprint arXiv:2507.07999},   year={2025} }

Acknowledgement

• Qwen2.5-VL: The base model we utilzed.
• VGR: The source of our SFT dataset.
• V* and VisDrone: The image source of our RL dataset.
• SA-1B: The image source of our TreeBench.
• LLaMA-Factory: The SFT codebase we utilized.
• EasyR1: The RL codebase we utilized.