This week, at #ICLR2026, we are presenting:
RAP: 3D Rasterization Augmented End-to-End Planning
👉 Project page w/ code: alan-lanfeng.github.io/RAP/
Collab with VITA lab from EPFL
@iclr-conf.bsky.social #Bench2Drive #navsim
🌍 WorldEngine: Towards the Era of Post-Training for Physical AI
🎯 A post-training framework for Physical AI that systematically addresses the long-tail safety-critical data scarcity problem in autonomous driving.
Github: github.com/OpenDriveLab...
Project Page: opendrivelab.com/WorldEngine/
🌍WorldEngine is one of the most exciting projects in AD in the past years!
It's a post-training framework tackling the scarcity of long-tail safety-critical scenarios by: mining -> 3DGS reconstruction and dynamic agents control w/ behavior world models -> RL post-training.
Blog, code and data are up
Congrats @vickykalogeiton.bsky.social! Well deserved!
This very nice paper provides some useful pushback against PRH.
To me science is like a damped pendulum, where we need to swing back and forth a few times before converging on truth.
So don't worry PRH fans, I'll be trying to swing us back out of the cave again soon!
Back to self-supervised basics (rotation prediction, colorization) in the era of MLLMs.
We discovered that simply interleaving SSL tasks with instruction tuning ones improves visual grounding of MLLMs, especially on vision-centric tasks.
Check it out 👇
1/n New paper - V-GIFT 🎁
Self-supervised tasks like rotation prediction or colorization were big in 2018.
Do they still matter?
Yes.
We turn them into visual instruction tuning data for MLLMs.
Result: models rely more on the image and perform better on vision tasks 👀
The NeurIPS submission site is open.
IDs are already > 650
🚨 arxiv.org/abs/2604.06129
PoM: A Linear-Time Replacement for Attention with the Polynomial Mixer
This paper is the result of doing a lab-wide hackathon on an idea I've had for some time. Probably the paper with the highest number of authors I've ever done.
It's a CVPR Findings 26.
Thread 🧵👇
Of course
All booked for #CVPR2026
So much looking forward to the papers, posters, workshops, people, chats … and mountains ⛰️
See you in Denver!
Great combo of cherry blossoms + sun this Easter in Sceaux.
🚨 Happy to announce CVPR@Paris'26 which will take place on June 1st in Paris. The goal of the event is to share a little bit of the conference before it happens. We will have poster sessions as well as several plenary talks by world-class speakers.
info: cvprinparis.github.io/CVPR2026InPa...
The DUNE encoder is a ViT distilled from multiple heterogenous teachers and is incredibly powerful, even outperforming the MASt3R on localization. We use it for navigation and manipulation.
@mbsariyildiz.bsky.social
@weinzaepfelp.bsky.social
@skamalas.bsky.social
@dlarlus.bsky.social
et al.
Today I launched blunder.clinic, a realistic daily #chess puzzle app!
Short 🧵
Within just 3 years, Victor's work and contributions led to 3 x NeurIPS, 2 x ICML, 1 x ICRA, 1 x preprint.
@vletzelter.bsky.social is a brilliant, yet humble researcher and a joy to work with.
We're super proud of his progress, wish him good luck and congratulate the lab that will have him ^^
Congrats Dr. @vletzelter.bsky.social for his work defended in front of a stellar jury and answering to so many questions that @rflamary.bsky.social decided a break at some point ^^
Fun fact: Victor had not 4, not 6, but *8* advisors over the span of his thesis and he handled them brilliantly
The #3DV2026 Keynote and Award Talk recordings are officially live! 🎥🍿
Revisit all the fantastic presentations from our insightful speakers and keep the 3D vision inspiration going!
See the links below⬇️
🚨New PhD in the house🚨
Congrats to Dr. Simon de Moreau for his PhD work on Active perception for nighttime scene understanding using vehicle lighting.
Incredible motivation and resilience to push ideas at the forefront of research and real world products simon.demoreau.fr
EurIPS will be in Paris this year and an official presentation venue! 🚀
Hey @3dvconf.bsky.social, we have some fun papers to present this week: a ridiculously simple strategy for LiDAR instance segmentation and a strong open-vocabulary perception LiDAR encoder #3DV2026
Thanks Eugene for making it.
I'm using the popularity of your content to lure X people into bluesky 😅
Naver AI has put the word "World" into "World Models", at least on a metropolis scale: A world model for Seoul using Naver Maps (the Seoul capital area is 25.6 million people, btw.).
seoul-world-model.github.io
arxiv.org/abs/2603.15583
@jnhwkim.bsky.social
To ensure compliance w peer-review policies, ICML has removed 795 reviews (1% of total) by reviewers who used LLMs when they explicitly agreed to not. Consequently, 497 papers (2% of all submissions) of these (reciprocal) reviewers have been desk rejected
Details in blog post 👇
An unsolicited guide to being a researcher: super instructive slides by @eugenevinitsky.bsky.social
He covers:
- different goals of a PhD student
- how to be a good collaborator (so useful)
- how to keep up with literature
- tracking your ideas & experiments
- stress & productivity
A 7-hour marathon interview with @saining.bsky.social: m.youtube.com/watch?v=rIwg...
Already saw a few fun bits from it, but can someone please do English audio for it?
cc @valeoai.bsky.social
Tried this out of curiosity early in the morning. I guess I can still work as slop-detector if research doesn’t work 😅
Thorough post by Nicholas Carlini with lots of insightful bits on how to choose research ideas to pursue, when to give them up or go on and how to write good papers nicholas.carlini.com/writing/2026...
