Update README.md
Browse files
README.md
CHANGED
|
@@ -1,8 +1,5 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
-
base_model:
|
| 4 |
-
- Qwen/Qwen3-4B-Thinking-2507
|
| 5 |
-
- apple/aimv2-large-patch14-448
|
| 6 |
pipeline_tag: image-text-to-text
|
| 7 |
library_name: transformers
|
| 8 |
---
|
|
@@ -11,6 +8,7 @@ library_name: transformers
|
|
| 11 |
<h1>AndesVL-4B-Thinking</h1>
|
| 12 |
<a href='https://arxiv.org/abs/2510.11496'><img src='https://img.shields.io/badge/arXiv-2510.11496-b31b1b.svg'></a>
|
| 13 |
<a href='https://huggingface.co/OPPOer'><img src='https://img.shields.io/badge/🤗%20HuggingFace-AndesVL-ffd21f.svg'></a>
|
|
|
|
| 14 |
</div>
|
| 15 |
AndesVL is a suite of mobile-optimized Multimodal Large Language Models (MLLMs) with **0.6B to 4B parameters**, built upon Qwen3's LLM and various visual encoders. Designed for efficient edge deployment, it achieves first-tier performance on diverse benchmarks, including those for text-rich tasks, reasoning tasks, Visual Question Answering (VQA), multi-image tasks, multilingual tasks, and GUI tasks. Its "1+N" LoRA architecture and QALFT framework facilitate efficient task adaptation and model compression, enabling a 6.7x peak decoding speedup and a 1.8 bits-per-weight compression ratio on mobile chips.
|
| 16 |
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
| 3 |
pipeline_tag: image-text-to-text
|
| 4 |
library_name: transformers
|
| 5 |
---
|
|
|
|
| 8 |
<h1>AndesVL-4B-Thinking</h1>
|
| 9 |
<a href='https://arxiv.org/abs/2510.11496'><img src='https://img.shields.io/badge/arXiv-2510.11496-b31b1b.svg'></a>
|
| 10 |
<a href='https://huggingface.co/OPPOer'><img src='https://img.shields.io/badge/🤗%20HuggingFace-AndesVL-ffd21f.svg'></a>
|
| 11 |
+
<a href='https://github.com/OPPO-Mente-Lab/AndesVL_Evaluation'><img src="https://img.shields.io/badge/GitHub-OPPOer-blue.svg?logo=github" alt="GitHub"></a>
|
| 12 |
</div>
|
| 13 |
AndesVL is a suite of mobile-optimized Multimodal Large Language Models (MLLMs) with **0.6B to 4B parameters**, built upon Qwen3's LLM and various visual encoders. Designed for efficient edge deployment, it achieves first-tier performance on diverse benchmarks, including those for text-rich tasks, reasoning tasks, Visual Question Answering (VQA), multi-image tasks, multilingual tasks, and GUI tasks. Its "1+N" LoRA architecture and QALFT framework facilitate efficient task adaptation and model compression, enabling a 6.7x peak decoding speedup and a 1.8 bits-per-weight compression ratio on mobile chips.
|
| 14 |
|