@HanwenXu6 Profile picture

Hanwen Xu

@HanwenXu6

PhD student at CSE, UW @uwcse

Joined March 2022
Similar User
Mike A. Merrill photo

@Mike_A_Merrill

Sheng Wang photo

@wangshengpkucn

Alexandru Coca photo

@alexcoca23

Yongchan Kwon photo

@yckwon0407

Wei Qiu photo

@weiqiu55

Xinming Tu photo

@TuXinming

Abe Hou photo

@abe_hou

Shitong Luo photo

@luost26

Ethan Weinberger photo

@efweinberger

Inna Lin photo

@iwylin

Bingbing Wen photo

@bingbingwen1

Binchi Zhang photo

@zhangbinchi3

Ian Covert photo

@ianccovert

Liyuan Liu (Lucas) photo

@LiyuanLucas

Qisheng Li photo

@QishengLi

Hanwen Xu Reposted

Analysis of o1-preview on medical benchmarks, comparing o1's native inference-time powers to prior work on Medprompt, a different approach to inference-time analysis. Assessment: o1 reaches new heights--and shows great promise for medical applications. tinyurl.com/9yspcpm6

Tweet Image 1

Hanwen Xu Reposted

At our next @StanfordHAI seminar, learn how specialized models are shaping the future of healthcare. @UW Asst. Prof. @wangshengpkucn will introduce recent works towards building multimodal biomedicine foundation models. Save your spot here: stanford.io/3BSrhjt

Tweet Image 1

Hanwen Xu Reposted

Excited to share OCTCube (arxiv.org/abs/2408.11227), the first *3D* foundation model for retinal OCT images. OCTCube is trained on 26k 3D OCT volumes encompassing 1.62 million 2D images. OCTCube achieved significantly best performance on 27 out of 29 tasks.

Tweet Image 1

Hanwen Xu Reposted

I think we may be approaching a tipping point in AI for medical imaging. Incredible progress at the @CVPR workshop on foundation models for medical imaging. Our teams were pleased to share work @MSFTResearch on #Biomeparse and #GigaPath to standing-room only audience.

Tweet Image 1

Hanwen Xu Reposted

🔬In latest @natBME paper, we introduce a new platform for clinician+AI collaboration: nuclei Most medical AI is one-size-fits-all. But different clinicians have diff use cases/requirements. Nuclei enables user to rapidly personalize AI via active QA nature.com/articles/s4155… 1/4


Hanwen Xu Reposted

With the fast-growing interest in multi-image understanding, where can we find a high-quality, comprehensive & robust benchmark? A group of talented students have created ᴍᴜɪʀʙᴇɴᴄʜ, a long-awaited multi-image benchmark, with 2.6k multi-choice Qs and over 10k images. (1/6)

Tweet Image 1

Hanwen Xu Reposted

🚀 Exciting news! @suinleelab, @mkaeberlein, and @weiqiu55's groundbreaking research has led to @NovosLabs' free non-invasive and state-of-the-art AI-driven biological age test. Now, anyone can gain accurate longevity insights through a simple questionnaire and the free NOVOS…


Hanwen Xu Reposted

#UWAllen’s @wangshengpkucn & @HanwenXu6 worked with @MSFTResearch & @Providence to create Prov-GigaPath, a whole-slide #AI model for digital pathology that achieves state-of-the-art performance on cancer subtyping and mutation prediction tasks. In @Nature: news.cs.washington.edu/2024/05/29/in-…


🔥🔥Super excited to share our GigaPath paper @Nature! Model: huggingface.co/prov-gigapath/… Code: github.com/prov-gigapath/… Really enjoy the journey towards ultimate patient care and clinical discovery! This is just the beginning!

Just out @Nature The 1st whole-slide digital pathology #AI foundation model pre-trained on large-scale real-world data, from over 1.3 billion images, 30,000 patients nature.com/articles/s4158… @hoifungpoon @HanwenXu6 @Microsoft @UW @naotous @MSFTResearch

Tweet Image 1
Tweet Image 2


Hanwen Xu Reposted

Generative #AI for medical images: incorporating synthetic images in diffusion models improved fairness, accuracy for underrepresented people nature.com/articles/s4159… @NatureMedicine @s0f1ra @sgowal @alan_karthi @GoogleDeepMind


Hanwen Xu Reposted

Honored to deliver a distinguished keynote at ISMB 2024! Join us in Montreal this July 12-16: iscb.org/ismb2024/progr…


Hanwen Xu Reposted

New breakthrough from Microsoft: 1-bit LLMs. New models that use ternary values (-1, 0, 1) instead of 16-bit. This makes them 2.7x faster, use 3.5x less GPU memory, and 71x less energy. Bitnet also matches or outperformed traditional models like LLaMA 3B.

Tweet Image 1

Hanwen Xu Reposted

❗With dense retrieval, the unit in which you segment a retrieval corpus (passage, sentence, etc) may impact performance by more than you thought! We introduce a novel retrieval unit, proposition, for dense retrieval. chentong0.github.io/factoid-wiki/ [1/7]

Tweet Image 1

Hanwen Xu Reposted

🤯Medprompt isn't just for medical queries. GPT-4 achieves 90.1% on MMLU with MedPrompt+, outperforming Gemini across the board! 🏆Dive into the promptbase on GitHub (github.com/microsoft/prom…) to unleash the power of advanced AI! #AI #GPT4 #Medprompt

Tweet Image 1

Today, we're sharing several updates on methods for steering foundation models aka.ms/AAo3s1z, including studies that build on our earlier #Medprompt work and a new resource on github with scripts & info #promptbase @MSFTResearch



Hanwen Xu Reposted

Thank you #MLCB2023 organizers @sara_mostafavi @david_a_knowles @anshulkundaje @suinleelab @QuonBio @james_y_zou and attendees for a vibrant meeting! It was great to have both virtual and in-person options and visit the @UW campus!

Tweet Image 1
Tweet Image 2
Tweet Image 3
Tweet Image 4

I’m a little surprised. I love working on-site but that really depends on people’s own style. More important to find the one that matches best.

In today's @Nature, a study shows that remote teams produce fewer breakthroughs! Disruptive science is more likely to come from groups working in close collaboration. 'Night science' discussions may require the kind of energy that's communicated in person. nature.com/articles/s4158…

Tweet Image 1


Hanwen Xu Reposted

Poisoning scientific knowledge using large language models. We use GPT to generate a fake scientific paper and mix this fake paper with 3 million real papers. We found that just one simple fake paper can disrupt scientific knowledge graphs! biorxiv.org/content/10.110…


Hanwen Xu Reposted

(1/6) Our deep dive into single cell foundation models, scBERT and scGPT - now on bioRxiv! biorxiv.org/content/10.110… w/ @nalinimsingh @abuen_dia @david_sontag @gaddyg


Loading...

Something went wrong.


Something went wrong.