Thank you Google Research for supporting our healthcare research! Honored to be one of the Google Research Scholars this year!

OpenVision is accepted by #ICCV2025 🥳🥳 Additionally, stay tuned for v2, arriving very soon with even greater efficiency and capability.

🥳Thrilled to share that our paper SPA: Efficient User-Preference Alignment against Uncertainty in Medical Image Segmentation has been accepted to #ICCV2025 🎉🎉 We present SPA — an efficient, user-friendly framework that aligns medical segmentation to clinician

🧵 1/ 🚀 Excited to share our latest work: Fractional Reasoning. We introduce a new way to continuously control the depth of reasoning and reflection in LLM for scaling test time compute, not just switch between “on” and “off” prompts. 💻 Website: shengliu66.github.io/fractreason/ #AI

📢 Women in #MICCAI Webinar: From Paper to Story: Crafting Effective Research Presentations - Best Practices from Oral Presentation Winner and Experienced Researchers 📅 Thursday, July 10, 2025 ⏲️ 4:30 PM PDT / 7:30 PM EDT / 11:30 PM UTC  📝Registration: us02web.zoom.us/webinar/regist…

[1/7] Excited to share our new survey on Latent Reasoning! The field is buzzing with methods—looping, recurrence, continuous thoughts—but how do they all relate? We saw a need for a unified conceptual map. 🧵 📄 Paper: arxiv.org/abs/2507.06203 💻 Github: github.com/multimodal-art…

A Survey on Latent Reasoning

A Survey of Latent Reasoning Nice overview on the emerging field of latent reasoning. Great read for AI devs. (bookmark it)

Models' hidden thoughts make a big impact. That's why "A Survey on Latent Reasoning" is a must-read. It explores how models reason in hidden states — Latent Chain-of-Thought, covering: - Higher-bandwidth latent reasoning - 2 key approaches: vertical vs. horizontal - Training

Most current language models think out loud, stuffing every thought into words. A typical token set holds about 40000 choices, which equals roughly 15 bits of data, just under 2 bytes. When a language model must pour every reasoning step through these tiny packets, complex

Top AI Papers of The Week (July 7 - 13): - H-Net - HIRAG - Kimi K2 - MemAgent - Adaptive Branching MCTS - A Survey on Latent Reasoning - What Has a Foundation Model Found? Read on for more:

Just arrived at #ICML25! If you're interested in multimodality, reasoning, and safety — let's connect. Additionally, I'll also be presenting two papers: 📄 What If We Recaption Billions of Web Images with LLaMA-3? 🗓 Tue, Jul 15 | 11 AM – 1:30 PM PDT 📍 East Hall A-B (#E-3305)

If you are at ICML2025, welcome to attend our workshop "Multi-modal Foundation Models and Large Language Models for Life Sciences" tomorrow (Jul 19 Saturday). The workshop features a stellar lineup of invited speakers, including Valentina Boeva Valentina Boeva, Manolis Kellis

🙌 We've released the full version of our paper, OpenVLThinker: Complex Vision-Language Reasoning via Iterative SFT-RL Cycles Our OpenVLThinker-v1.2 is trained through three lightweight SFT → RL cycles, where SFT first “highlights” reasoning behaviors and RL then explores and

🚀 Excited to share GPT-Image-Edit-1.5M — our new large-scale, high-quality, fully open image editing dataset for the research community! (1/n)

Seems like flux-kontext can be fine-tuned

It should have been like this all along, lumina2(with gemma2) OmniGen2(MLLM) is the right choice, it's time to replace T5.

🚀 One of our group’s largest efforts in building high-quality image editing datasets: arxiv.org/abs/2507.21033 Using the latest GPT-Image API, we unified 3 top datasets—OmniEdit, HQ-Edit, and UltraEdit—into ~1.5M editing pairs with incredible visual quality and

BraTS, LiTS, and KiTS have deeply advanced cancer research. Now, PanTS joins this family of open datasets! I’m truly grateful for the community efforts that make these valuable resources available to everyone. github.com/MrGiovanni/Pan…

We are excited to share the Call for Papers for the NeurIPS 2025 2nd Workshop on Multi-modal Foundation Models and Large Language Models for Life Sciences. We welcome paper submissions that are up to 4 pages in length (excluding references) formatted using NeurIPS 2025 paper