Xiao

We are looking for PhD candidates for Sep 2025 or Jan 2026 admission to explore exciting technologies in neuro AI and neuro-feedback VR! 🧠⚡️

We should always think about what does it mean for doing research than heavy emphasis on metrics. Even for ML algorithms, the wrong cost function easily misguides the outcomes.

If the writing of your paper is not even finished, please don't proceed with the submission for review. It is wasting the time of the reviewers.

Excited to share our #CVPR2026 paper "MedCLIPSeg: Probabilistic Vision-Language Adaptation for Data-Efficient and Generalizable Medical Image Segmentation" by brilliant PhD trainee Taha Koleilat and co-authors! Adding uncertainty in CLIP-based segmentation tahakoleilat.github.io/MedCLIPSeg/

MICCAI deadline done! 😎

@benjaminbardou You do really beautiful works!

Studies of Artificial Landscapes Part of a research project on new forms created by artificial imagination. Just as plein-air painters like Eugène Boudin were captivated by the shape and color that a landscape might take at certain times of day, the aim here is to capture the geography of latent space. Sketching thus appears to be the technique best suited to capturing the infinitely sensitive fluctuations of this new imaginary space.

@gadevenyi Still waiting

I just got a call from the Quebec family doctor list 11 years after enrollment! ... to update my file.

Great to know that two of our papers on biomedical vision language models have been accepted at CVPR2026! Congratulations to our brilliant trainees! 👏🎊

GM if you love computer generative art!

CIHR funding rate: 13.6%. Early 2000s: 30%+. Canada’s new $1.7B Impact+ program recruits talent — but without increased Tri-Council base funding, we risk further strain on an already stretched system. We’re calling for $1B over 5 years. Support here: tinyurl.com/33c5av46

When I went through my researchgate inbox today, I saw the kind message from Dr Karl Zilles again, who I never met in person. It really made my day many years ago. In academia, we are constantly being assessed. Don't forget to be kind and encouraging as well.

At the same time, the number of @icmlconf submissions went from 12k in 2025 to over 30k this year, and 50 published papers at @NeurIPSConf were found to contain hallucinated references. We're running heads down into a severe publishing crisis in AI.

I keep running into empty GitHub repos (also at times promised data repos) from published papers in top venues. Open sourcing the code is not mandatory, but if it is part of the commitment expressed during review, it should be followed through

The three-body problem in physics it involves predicting the motion of three celestial objects interacting under gravity, this system is known for its chaotic and unpredictable nature.

@Meta why are you shutting down supernatural?🫠

Wow! Looking forward to this!

It is tiring to see all the AI-generated images and videos all over the place, including LinkedIn. The shock factor fades quickly when supposed creativity becomes generic and empty.

A human-AI collaboration design can greatly benefit medical image segmentation for both the AI system and clinicians in accuracy, reliability, and skill training. Check out our latest publication in Virtual Reality! link.springer.com/article/10.100… youtu.be/DOCJS23zfwU?si…

So cool!

Happy holidays from the Health-X Lab! 🎄❄️

One of my favorite moments from Yejin Choi’s NeurIPS keynote was her point as follows: "it looks like a minor detail, but one thing I learned since joining and spending time at NVIDIA is that all these, like, minor details, implementation details matter a lot" -- I think this is exactly the point that theory people often undervalue when it comes to empirical work.