10 16 13

Yuan Liu

YuanLiuuuuuu

https://yuanliuuuuuu.github.io/

AI & ML interests

Large Multimodal Model, Large Language Model

Recent Activity

upvoted a paper about 2 months ago

Lost in Embeddings: Information Loss in Vision-Language Models

new activity about 2 months ago

tencent/POINTS-Reader:AttributeError: 'POINTSV15ChatConfig' object has no attribute 'hidden_size'

new activity 2 months ago

tencent/POINTS-Reader:Adding a demo Colab notebook. ↗️

View all activity

Organizations

upvoted a paper about 2 months ago

Lost in Embeddings: Information Loss in Vision-Language Models

Paper • 2509.11986 • Published Sep 15 • 27

New activity in tencent/POINTS-Reader about 2 months ago

AttributeError: 'POINTSV15ChatConfig' object has no attribute 'hidden_size'

#2 opened 2 months ago by

nilecui

New activity in tencent/POINTS-Reader 2 months ago

Adding a demo Colab notebook. ↗️

#3 opened 2 months ago by

prithivMLmods

New activity in prithivMLmods/POINTS-Reader-OCR 2 months ago

Latex formula enclosed in $xxx$ can not be rendered correctly

#8 opened 2 months ago by

YuanLiuuuuuu

updated a model 2 months ago

tencent/POINTS-Reader

Image-Text-to-Text • 4B • Updated Sep 12 • 256 • 97

replied to prithivMLmods's post 2 months ago

By the way, would you be open to exploring the possibility of accelerating this demo using SGLang, following the guide available here? I believe it could offer some nice performance improvements.

reacted to prithivMLmods's post with 🚀👍 2 months ago

Post

3122

The POINTS-Reader, a vision-language model for end-to-end document conversion, is a powerful, distillation-free Vision-Language Model that sets new SoTA benchmarks. The demo is now available on HF (Extraction, Preview, Documentation). The input consists of a fixed prompt and a document image, while the output contains only a string (the text extracted from the document image). 🔥🤗

✦ Space/App: prithivMLmods/POINTS-Reader-OCR
✦ Model: tencent/POINTS-Reader
✦ Paper: https://arxiv.org/pdf/2509.01215

🤗 The app is done and ready to go brrrr with zero GPU. Thankyou @merve

.
.
.
To know more about it, visit the app page or the respective model page!!