Asst Prof in @NUSingapore / Former senior research fellow @WEISS_UCL, Ph.D. CUHK. Multi-modal AI in Healthcare; Surgical Data Science; Surgical Robotics
Excited to introduce our recent work #SurgVLM, a universal vision-language foundation model for surgical intelligence, aiming to support cognitive assistance, facilitate surgical education, task automation. Welcome to check!😀
🔗https://t.co/53f6MXemGV
🔗https://t.co/P9OFRhVCXC
Introducing OneContext. I built it for myself but now I can’t work without it, so it felt wrong not to share.
OneContext is an Agent Self-Managed Context Layer across different sessions, devices, and coding agents (Codex / Claude Code).
How it works:
1. Open Claude Code/Codex inside OneContext as usual, it automatically manages your context and history into a persistent context layer.
2. Start a new agent under the same context, it remembers everything about your project.
3. Share the context via link, anyone can continue building on the exact same shared context.
Install with: npm i -g onecontext-ai
And open with: onecontext
Give it a try!
Thanks @omarsar0 for the great summarization of our work "KNOWLEDGE or REASONING ? A Close Look at How LLMs Think Across Domains"!
How can we build more reliable LLMs? 🤔
We focus on ensuring not just accurate final answers, but also high-quality reasoning 💡 & knowledge utilization 🧠 at every single step!
Here’s what we found & advocate for:
➡️ Evaluate step-wise knowledge (Knowledge Index) & reasoning (InfoGain), not just the final output.
➡️ SFT boosts knowledge (vital for domains like medicine!) but can make reasoning verbose, sometimes reducing its quality.
➡️ RL enhances reasoning quality & can prune incorrect knowledge from reasoning paths, improving overall knowledge correctness.
Paper: https://t.co/hNyjhrAyHr
Project: https://t.co/uDHVFeC4bD
Code: https://t.co/8YDANBOpmK
Excited to introduce our recent work #SurgVLM, a universal vision-language foundation model for surgical intelligence, aiming to support cognitive assistance, facilitate surgical education, task automation. Welcome to check!😀
🔗https://t.co/53f6MXemGV
🔗https://t.co/P9OFRhVCXC
Large Vision-Language Models (#VLMs) have shown promising performance in various natural domain applications. While for medical, especially the surgical domain, VLMs are still largely underexplored. We hope this work can help alleviate this gap.
4/
We are excited to share our latest work: "ReSurgSAM2: Referring Segment Anything in Surgical Video via Credible Long-term Tracking", early accepted at MICCAI 2025 (top 9%)🚀
arXiv: https://t.co/G6whZRUNRm
Huggingface: https://t.co/kJhdkqVpis
#SurgicalAI#SugicalDataScience#SAM2
Excited to share that we will host the 1st Workshop on Collaborative Intelligence and Autonomy in Image Guided Surgery (#COLAS) at #MICCAI2025!
🥤Welcome to submit your work! -- deadline: July 6, 2025
🥤More info: https://t.co/gwjg0vGTJ9
Excited to share our work: "MedAgent-Pro: Towards Evidence-Based Multi-Modal Medical Diagnosis via Reasoning Agentic Workflow":
Paper: https://t.co/xNIccrdfQX
Code: https://t.co/uhamZgihf0
Huge thanks to @JinYueming , @JundeMorsenWu , and @lowchanghan99 for their effort!🎉🎉
I am thrilled to share our latest work: “SurgRAW: Multi-Agent Chain-of-Thought Reasoning for Surgical Intelligence”
Arxiv: https://t.co/tHaxJaRi8x
Special thanks to Prof @JinYueming and @AI_Ziyue and all collaborators who made this work possible.
🔥Introducing Open-Janus: a fully open-source and reproducible training pipeline for Janus.
Janus—@DeepSeek_AI’s powerful vision-language model—unifies multimodal understanding and generation in one framework.
We’re strong believers that this is a promising path toward truly generalized vision-language intelligence.
But while the model weights were released, the training code wasn’t.🫠.
So we rebuilt it from scratch. And now it’s open for everyone!🥳
Check it out 👉 https://t.co/yENfFsiXtQ
We've already achieved image understanding performance very close to DeepSeek’s Janus, and rapidly closing the gap on the others.
Contributions welcome!
Grateful for the work by Fang Zheng @meaw041524005 and the guidance of @JinYueming, which made it possible.
#MultimodalAI #VisionLanguage #OpenSource #LLM #Janus #DeepLearning
📢Happy to share our MedSAM-2, a novel way for segmenting medical images as videos based on SAM-2.
STRONGEST model for promptable medical image segmentation for both 2D and 3D images (+500 GH stars)
joint work with collaborators from @UniofOxford and NUS
a thread 🧵
1/n
Excited to share our work: "Surgical SAM 2: Real-time Segment Anything in Surgical Video by Efficient Frame Pruning" at the AIM-FM workshop, NeurIPS 2024. We look forward to continuing our exploration and contributions to the research community.
#NeurIPS2024#SurgicalAI