SenseNova-SI-1.2-InternVL3-8B

EN | 中文

SenseNova-SI: Scaling Spatial Intelligence with Multimodal Foundation Models

<a href="https://github.com/OpenSenseNova/SenseNova-SI" target="_blank"> <img alt="Code" src="https://img.shields.io/badge/SenseNova_SI-Code-100000?style=flat-square&logo=github&logoColor=white" height="20" /> </a> <a href="https://arxiv.org/abs/2511.13719" target="_blank"> <img alt="arXiv" src="https://img.shields.io/badge/arXiv-SenseNova_SI-red?logo=arxiv" height="20" /> </a> <a href="https://github.com/EvolvingLMMs-Lab/EASI" target="_blank"> <img alt="Code" src="https://img.shields.io/badge/EASI-Code-100000?style=flat-square&logo=github&logoColor=white" height="20" /> </a> <a href="https://easi.lmms-lab.com/leaderboard" target="_blank"> <img alt="Leaderboard" src="https://img.shields.io/badge/%F0%9F%A4%97%20_EASI-Leaderboard-ffc107?color=ffc107&logoColor=white" height="20" /> </a>

Overview

Despite remarkable progress, multimodal foundation models still exhibit surprising deficiencies in spatial intelligence. In this work, we explore scaling up multimodal foundation models to cultivate spatial intelligence within the SenseNova-SI family, built upon established multimodal foundations including visual understanding models (i.e., Qwen3-VL and InternVL3) and unified understanding and generation models (i.e., Bagel). We take a principled approach to constructing high-performing and robust spatial intelligence by systematically curating SenseNova-SI-8M: eight million diverse data samples under a rigorous taxonomy of spatial capabilities. SenseNova-SI demonstrates unprecedented performance across a broad range of spatial intelligence benchmarks, while maintaining strong general multimodal understanding. More importantly, we analyze the impact of data scaling, discuss early signs of emergent generalization capabilities enabled by diverse data training, analyze the risk of overfitting and language shortcuts, present a preliminary study on spatial chain-of-thought reasoning, and validate the potential downstream application. SenseNova-SI is an ongoing project, and this report will be updated continuously. All newly trained multimodal foundation models are publicly released to facilitate further research in this direction. In the future, SenseNova-SI will be integrated with larger-scale in-house models.

Release Information

Currently, we build SenseNova-SI upon popular open-source foundation models to maximize compatibility with existing research pipelines. In this release, we present SenseNova-SI-1.2-InternVL3-8B, SenseNova-SI-1.1-Qwen2.5-VL-3B, SenseNova-SI-1.1-Qwen2.5-VL-7B, and SenseNova-SI-1.1-Qwen3-VL-8B, of which SenseNova-SI-1.2-InternVL3-8B achieve state-of-the-art performance among open-source models of comparable size across eight recent spatial intelligence benchmarks: VSI, MMSI, MindCube, ViewSpatial, SITE, BLINK, 3DSRBench, EmbSpatial-Bench.

<table> <thead> <tr> <th>Model</th> <th>VSI</th> <th>MMSI</th> <th>MindCube-Tiny</th> <th>ViewSpatial</th> <th>SITE</th> <th>BLINK</th> <th>3DSRBench</th> <th>EmbSpatial-Bench</th> </tr> </thead> <tbody> <tr style="background:#F2F0EF;font-weight:700;text-align:center;"> <td colspan="9"><em>Open-source Models (~2B)</em></td> </tr> <tr> <td>InternVL3-2B</td><td>32.9</td><td>26.5</td><td>37.5</td><td>32.5</td><td>30.0</td><td>50.8</td><td>47.7</td><td>60.1</td> </tr> <tr> <td>Qwen3-VL-2B-Instruct</td><td>50.3</td><td>28.9</td><td>34.5</td><td>36.9</td><td>35.6</td><td>53.2</td><td>47.5</td><td>70.1</td> </tr> <tr> <td>MindCube-3B-RawQA-SFT</td><td>17.2</td><td>1.7</td><td>51.7</td><td>24.1</td><td>6.3</td><td>35.1</td><td>2.8</td><td>37.0</td> </tr> <tr> <td>SpatialLadder-3B</td><td>44.8</td><td>27.4</td><td>43.4</td><td>39.8</td><td>27.9</td><td>43.0</td><td>42.8</td><td>58.2</td> </tr> <tr> <td>SpatialMLLM-4B</td><td>46.3</td><td>26.1</td><td>33.4</td><td>34.6</td><td>18.0</td><td>40.5</td><td>36.2</td><td>50.0</td> </tr> <tr> <td>VST-3B-SFT</td><td>57.9</td><td>30.2</td><td>35.9</td><td>52.8</td><td>35.8</td><td>58.8</td><td>54.1</td><td>69.0</td> </tr> <tr> <td>Cambrian-S-3B</td><td>57.3</td><td>25.2</td><td>32.5</td><td>39.0</td><td>28.3</td><td>37.7</td><td>50.9</td><td>63.5</td> </tr> <tr style="background:#F2F0EF;font-weight:700;text-align:center;"> <td colspan="9"><em>Open-source Models (~8B)</em></td> </tr> <tr> <td>InternVL3-8B</td><td>42.1</td><td>28.0</td><td>41.5</td><td>38.6</td><td>41.1</td><td>53.5</td><td>44.3</td><td>76.4</td> </tr> <tr> <td>Qwen3-VL-8B-Instruct</td><td>57.9</td><td>31.1</td><td>29.4</td><td>42.2</td><td>45.8</td><td>66.7</td><td>53.9</td><td>77.7</td> </tr> <tr> <td>BAGEL-7B-MoT</td><td>31.4</td><td>31.0</td><td>34.7</td><td>41.3</td><td>37.0</td><td>63.7</td><td>50.2</td><td>73.1</td> </tr> <tr> <td>SpaceR-7B</td><td>41.5</td><td>27.4</td><td>37.9</td><td>35.8</td><td>34.2</td><td>49.6</td><td>40.5</td><td>66.9</td> </tr> <tr> <td>ViLaSR-7B</td><td>44.6</td><td>30.2</td><td>35.1</td><td>35.7</td><td>38.7</td><td>51.4</td><td>46.6</td><td>67.3</td> </tr> <tr> <td>VST-7B-SFT</td><td>60.6</td><td>32.0</td><td>39.7</td><td>50.5</td><td>39.6</td><td>61.9</td><td>54.6</td><td>73.7</td> </tr> <tr> <td>Cambrian-S-7B</td><td>67.5</td><td>25.8</td><td>39.6</td><td>40.9</td><td>33.0</td><td>37.9</td><td>54.8</td><td>72.8</td> </tr> <tr> <td><strong>SenseNova-SI-1.2-InternVL3-8B</strong></td> <td><strong>69.6</strong></td> <td><strong>42.6</strong></td> <td><strong>89.0</strong></td> <td><strong>58.8</strong></td> <td><strong>49.0</strong></td> <td><strong>69.4</strong></td> <td><strong>60.1</strong></td> <td><strong>77.7</strong></td> </tr> <tr style="background:#F2F0EF;color:#6b7280;font-weight:600;text-align:center;"> <td colspan="9"><em>Proprietary Models</em></td> </tr> <tr style="color:#6b7280;"> <td>Gemini-2.5-pro-2025-06</td><td>53.5</td><td>38.0</td><td>57.6</td><td>46.0</td><td>57.0</td><td>73.5</td><td>59.3</td><td>78.9</td> </tr> <tr style="color:#6b7280;"> <td>Grok-4-2025-07-09</td><td>47.9</td><td>37.8</td><td>63.5</td><td>43.2</td><td>47.0</td><td>56.4</td><td>54.9</td><td>75.7</td> </tr> <tr style="color:#6b7280;"> <td>GPT-5-2025-08-07</td><td>55.0</td><td>41.8</td><td>56.3</td><td>45.5</td><td>61.8</td><td>68.0</td><td>60.3</td><td>81.6</td> </tr> </tbody> </table>

🛠️ QuickStart

Installation

We recommend using uv to manage the environment.

uv installation guide: https://docs.astral.sh/uv/getting-started/installation/#installing-uv

git clone [email protected]:OpenSenseNova/SenseNova-SI.git
cd SenseNova-SI/
uv sync --extra cu124 # or one of [cu118|cu121|cu124|cu126|cu128|cu129], depending on your CUDA version
uv sync
source .venv/bin/activate

Hello World

A simple image-free test to verify environment setup and download the model.

python example.py \
  --question "Hello" \
  --model_path sensenova/SenseNova-SI-1.2-InternVL3-8B

Examples

Example 1

This example is from the Pos-Obj-Obj subset of MMSI-Bench:

python example.py \
  --image_paths examples/Q1_1.png examples/Q1_2.png \
  --question "<image><image>
You are standing in front of the dice pattern and observing it. Where is the desk lamp approximately located relative to you?
Options: A: 90 degrees counterclockwise, B: 90 degrees clockwise, C: 135 degrees counterclockwise, D: 135 degrees clockwise" \
  --model_path sensenova/SenseNova-SI-1.2-InternVL3-8B 
# --model_path OpenGVLab/InternVL3-8B 

<!-- Example 1 --> <details open> <summary><strong>Details of Example 1</strong></summary> <p><strong>Q:</strong>You are standing in front of the dice pattern and observing it. Where is the desk lamp approximately located relative to you? Options: A: 90 degrees counterclockwise, B: 90 degrees clockwise, C: 135 degrees counterclockwise, D: 135 degrees clockwise</p> <table> <tr> <td align="center" width="50%" style="padding:4px;"> <img src="./examples/Q1_1.png" alt="First image" width="100%"> </td> <td align="center" width="50%" style="padding:4px;"> <img src="./examples/Q1_2.png" alt="Second image" width="100%"> </td> </tr> </table> <p><strong>GT: C</strong></p> </details>

Example 2

This example is from the Rotation subset of MindCube:

python example.py \
  --image_paths examples/Q2_1.png examples/Q2_2.png \
  --question "<image><image>
Based on these two views showing the same scene: in which direction did I move from the first view to the second view?
A. Directly left B. Directly right C. Diagonally forward and right D. Diagonally forward and left" \
  --model_path sensenova/SenseNova-SI-1.2-InternVL3-8B 
# --model_path OpenGVLab/InternVL3-8B

<!-- Example 2 --> <details open> <summary><strong>Details of Example 2</strong></summary> <p><strong>Q:</strong> Based on these two views showing the same scene: in which direction did I move from the first view to the second view? Directly left B. Directly right C. Diagonally forward and right D. Diagonally forward and left</p> <table> <tr> <td align="center" width="50%" style="padding:4px;"> <img src="./examples/Q2_1.png" alt="First image" width="100%"> </td> <td align="center" width="50%" style="padding:4px;"> <img src="./examples/Q2_2.png" alt="Second image" width="100%"> </td> </tr> </table> <p><strong>GT: D</strong></p> </details>

Test Multiple Questions in a Single Run

Prepare a file similar to examples/examples.jsonl, where each line represents a single question.

The model is loaded once and processes questions sequentially. The questions remain independent of each other.

For more details on the jsonl format, refer to the documentation for Single-Image Data and Multi-Image Data.

python example.py \
  --jsonl_path examples/examples.jsonl \
  --model_path sensenova/SenseNova-SI-1.2-InternVL3-8B 
# --model_path OpenGVLab/InternVL3-8B 

Evaluation

To reproduce the benchmark results above, please refer to EASI to evaluate SenseNova-SI on mainstream spatial intelligence benchmarks.

🖊️ Citation

@InProceedings{sensenova-si,
  title = {Scaling Spatial Intelligence with Multimodal Foundation Models},
  author = {Cai, Zhongang and Wang, Ruisi and Gu, Chenyang and Pu, Fanyi and Xu, Junxiang and Wang, Yubo and Yin, Wanqi and Yang, Zhitao and Wei, Chen and Sun, Qingping and Zhou, Tongxi and Li, Jiaqi and Pang, Hui En and Qian, Oscar and Wei, Yukun and Lin, Zhiqian and Shi, Xuanke and Deng, Kewang and Han, Xiaoyang and Chen, Zukai and Fan, Xiangyu and Deng, Hanming and Lu, Lewei and Pan, Liang and Li, Bo and Liu, Ziwei and Wang, Quan and Lin, Dahua and Yang, Lei},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year = {2026}
}