Kun Xiang commited on
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,54 @@
|
|
| 1 |
-
---
|
| 2 |
-
|
| 3 |
-
--
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
task_categories:
|
| 3 |
+
- text-generation
|
| 4 |
+
size_categories:
|
| 5 |
+
- 157K
|
| 6 |
+
license: apache-2.0
|
| 7 |
+
---
|
| 8 |
+
|
| 9 |
+
# Model Card for AtomThink-EMOVA-8B
|
| 10 |
+
|
| 11 |
+
The model is post-trained based on EMOVA-8B and the AtomThink framework, and can be used to solve complex multimodal mathematical problems.
|
| 12 |
+
|
| 13 |
+
Comparison of accuracy with state-of-the-art methods on MathVista and MathVerse:
|
| 14 |
+
|
| 15 |
+
| **Model** | **Inference** | **General** | **Math** | **Total** | **TL** | **TD** | **VI** | **VD** | **VO** | **Total** |
|
| 16 |
+
|-----------------------|---------------|-------------|----------|-----------|----------|----------|----------|----------|----------|-----------|
|
| 17 |
+
| Random Choice | - | - | - | 17.9 | 12.4 | 12.4 | 12.4 | 12.4 | 12.4 | 12.4 |
|
| 18 |
+
| Human | - | - | - | - | 70.9 | 71.2 | 61.4 | 68.3 | 66.7 | 66.7 |
|
| 19 |
+
| OpenAI o1 | Slow Think* | - | - | 73.9 | - | - | - | - | - | - |
|
| 20 |
+
| GPT-4o | CoT | - | - | 63.8 | - | - | - | - | - | - |
|
| 21 |
+
| GPT-4V | CoT | - | - | 49.9 | 56.6 | 63.1 | 51.4 | 50.8 | 50.3 | 54.4 |
|
| 22 |
+
| LLaVA-NeXT-34B | Direct | - | - | 46.5 | 25.5 | 33.8 | 23.5 | 20.3 | 15.7 | 23.8 |
|
| 23 |
+
| InternLM-XComposer2 | Direct | - | - | 57.6 | 17.0 | 22.3 | 15.7 | 16.4 | 11.0 | 16.5 |
|
| 24 |
+
| Qwen-VL-Plus | Direct | - | - | 43.3 | 11.1 | 15.7 | 9.0 | 13.0 | 10.0 | 11.8 |
|
| 25 |
+
| LLaVA-1.5-13B | Direct | - | - | 27.6 | 15.2 | 19.4 | 16.8 | 15.2 | 11.3 | 15.6 |
|
| 26 |
+
| G-LLaVA-7B | Direct | - | - | 53.4 | 20.7 | 20.9 | 17.2 | 14.6 | 9.4 | 16.6 |
|
| 27 |
+
| MAVIS-7B | Direct | - | - | - | 29.1 | 41.4 | 27.4 | 24.9 | 14.6 | 27.5 |
|
| 28 |
+
| LLaVA-Llama3-8B | Direct | 34.1 | 25.6 | 29.5 | 16.0 | 19.3 | 16.4 | 13.1 | 15.0 | 15.9 |
|
| 29 |
+
| EMOVA-8B-200k | Direct | 52.4 | 51.1 | 51.7 | 34.4 | 39.0 | 33.4 | 30.1 | 23.5 | 32.1 |
|
| 30 |
+
| EMOVA w/. Formatted | CoT | 30.9 | 31.3 | 31.1 | 26.5 | 36.5 | 25.3 | 20.4 | 19.8 | 25.7 |
|
| 31 |
+
| AtomThink-EMOVA | Direct | 53.9 | 52.4 | 53.1 | 33.6 | 39.0 | 33.8 | 28.0 | 24.4 | 31.8 |
|
| 32 |
+
| AtomThink-EMOVA | Quick Think | 48.7 | **54.4** | **51.8** | **36.5** | **42.4** | **34.1** | **32.9** | **29.7** | **35.1** |
|
| 33 |
+
| AtomThink-EMOVA | Slow Think | 48.9 | **57.0** | **53.3** | **42.1** | **51.5** | **39.0** | **36.7** | **33.1** | **40.5** |
|
| 34 |
+
|
| 35 |
+
|
| 36 |
+
# Citation
|
| 37 |
+
If you use this dataset in your research, please cite:
|
| 38 |
+
```text
|
| 39 |
+
@article{xiang2024atomthink,
|
| 40 |
+
title={AtomThink: A Slow Thinking Framework for Multimodal Mathematical Reasoning},
|
| 41 |
+
author={Xiang, Kun and Liu, Zhili and Jiang, Zihao and Nie, Yunshuang and Huang, Runhui and Fan, Haoxiang and Li, Hanhui and Huang, Weiran and Zeng, Yihan and Han, Jianhua and others},
|
| 42 |
+
journal={arXiv preprint arXiv:2411.11930},
|
| 43 |
+
year={2024}
|
| 44 |
+
}
|
| 45 |
+
@article{chen2024emova,
|
| 46 |
+
title={Emova: Empowering language models to see, hear and speak with vivid emotions},
|
| 47 |
+
author={Chen, Kai and Gou, Yunhao and Huang, Runhui and Liu, Zhili and Tan, Daxin and Xu, Jing and Wang, Chunwei and Zhu, Yi and Zeng, Yihan and Yang, Kuo and others},
|
| 48 |
+
journal={arXiv preprint arXiv:2409.18042},
|
| 49 |
+
year={2024}
|
| 50 |
+
}
|
| 51 |
+
```
|
| 52 |
+
|
| 53 |
+
# License
|
| 54 |
+
The checkpoint is released under the Apache 2.0 license. Please ensure proper attribution when using this dataset.
|