Hugging Face's logo

Quinn777
/
AtomThink-LLaVA-Llama3-8B

Safetensors
llava_llama
Model card Files Community
Quinn777 commited on
Commit
87e9936
·
verified ·
1 Parent(s): 51fd4a0

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +48 -3
README.md CHANGED
@@ -1,3 +1,48 @@
1
- ---
2
- license: apache-2.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
+ license: apache-2.0
3
+ ---
4
+
5
+ # Model Card for AtomThink-LLaVA-Llama3-8B
6
+
7
+ The model is post-trained based on LLaVA-Llama3-8B and the AtomThink framework, and can be used to solve complex multimodal mathematical problems.
8
+
9
+ # Comparison of accuracy with state-of-the-art methods on MathVista and MathVerse:
10
+ | **Model** | **Inference** | **General** | **Math** | **Total** | **TL** | **TD** | **VI** | **VD** | **VO** | **Total** |
11
+ |-----------------------|---------------|-------------|----------|-----------|----------|----------|----------|----------|----------|-----------|
12
+ | Random Choice | - | - | - | 17.9 | 12.4 | 12.4 | 12.4 | 12.4 | 12.4 | 12.4 |
13
+ | Human | - | - | - | - | 70.9 | 71.2 | 61.4 | 68.3 | 66.7 | 66.7 |
14
+ | OpenAI o1 | Slow Think* | - | - | 73.9 | - | - | - | - | - | - |
15
+ | GPT-4o | CoT | - | - | 63.8 | - | - | - | - | - | - |
16
+ | GPT-4V | CoT | - | - | 49.9 | 56.6 | 63.1 | 51.4 | 50.8 | 50.3 | 54.4 |
17
+ | LLaVA-NeXT-34B | Direct | - | - | 46.5 | 25.5 | 33.8 | 23.5 | 20.3 | 15.7 | 23.8 |
18
+ | InternLM-XComposer2 | Direct | - | - | 57.6 | 17.0 | 22.3 | 15.7 | 16.4 | 11.0 | 16.5 |
19
+ | Qwen-VL-Plus | Direct | - | - | 43.3 | 11.1 | 15.7 | 9.0 | 13.0 | 10.0 | 11.8 |
20
+ | LLaVA-1.5-13B | Direct | - | - | 27.6 | 15.2 | 19.4 | 16.8 | 15.2 | 11.3 | 15.6 |
21
+ | G-LLaVA-7B | Direct | - | - | 53.4 | 20.7 | 20.9 | 17.2 | 14.6 | 9.4 | 16.6 |
22
+ | MAVIS-7B | Direct | - | - | - | 29.1 | 41.4 | 27.4 | 24.9 | 14.6 | 27.5 |
23
+ | LLaVA-Llama3-8B | Direct | 34.1 | 25.6 | 29.5 | 16.0 | 19.3 | 16.4 | 13.1 | 15.0 | 15.9 |
24
+ | LLaVA w/. Formatted | CoT | 30.2 | 22.9 | 26.3 | 14.3 | 18.4 | 15.7 | 10.0 | 7.7 | 13.2 |
25
+ | AtomThink-LLaVA | Direct | 34.4 | 27.2 | 30.5 | 16.0 | 19.3 | 16.2 | 13.1 | 15.0 | 15.9 |
26
+ | AtomThink-LLaVA | Quick Think | **36.9** | **37.0** | **36.6** | **22.2** | **26.6** | **24.1** | **20.9** | **17.9** | **22.4** |
27
+ | AtomThink-LLaVA | Slow Think | **36.5** | **41.3** | **39.1** | **36.1** | **42.4** | **30.0** | **36.8** | **28.6** | **34.7** |
28
+
29
+ # Citation
30
+ If you use this dataset in your research, please cite:
31
+ ```text
32
+ @article{xiang2024atomthink,
33
+ title={AtomThink: A Slow Thinking Framework for Multimodal Mathematical Reasoning},
34
+ 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},
35
+ journal={arXiv preprint arXiv:2411.11930},
36
+ year={2024}
37
+ }
38
+ @article{liu2024visual,
39
+ title={Visual instruction tuning},
40
+ author={Liu, Haotian and Li, Chunyuan and Wu, Qingyang and Lee, Yong Jae},
41
+ journal={Advances in neural information processing systems},
42
+ volume={36},
43
+ year={2024}
44
+ }
45
+ ```
46
+
47
+ # License
48
+ The checkpoint is released under the Apache 2.0 license. Please ensure proper attribution when using this checkpoint.