I received a review like this five years ago. It’s probably the right time now to share it with everyone who wrote or got random discouraging reviews from ICML/ACL.

@levelsio Sadly this is often the way these things work - contributions from small independent researchers get lost in the noise of big tech companies and prestigious universities.

NeurIPS D&B track in a nutshell: (1) An LLM-generated benchmark dataset (2) used to test performance of LLMs (3) evaluated via LLM-as-a-judge

🤔 Have OpenAI o3, Gemini 2.5, Claude 3.7 formed an internal world model to understand the physical world, or just align pixels with words? We introduce WM-ABench, the first systematic evaluation of VLMs as world models. Using a cognitively-inspired framework, we test 15 SOTA

VLMs are often used for planning across different world modelling scenarios. Check out this recent work by Qiyue Gao which helps highlight some of their limitations and strength.

As ICML 2025 is approaching, it's time to reheat this banger

Classic Unity/Unreal = hand-built assets + rigid solvers; neural game engines like Mirage can run interactive sandboxes from a prompt. 1993 GPUs solved graphics; physics and world assets stayed on the CPU. 2025 GPUs run giant diffusion models that have learnt the geometry,

Elon Musk How about AI video _game_ generation 🕹️

I have been long arguing that a world model is NOT about generating videos, but IS about simulating all possibilities of the world to serve as a sandbox for general-purpose reasoning via thought-experiments. This paper proposes an architecture toward that arxiv.org/abs/2507.05169

Some critical reviews and clarifications on different perspectives of world models. 🔥🌶️ Stay tuned for more on PAN — its position on the roadmap towards next-level intelligence, strong results, and open-sources❗️🧠

🚀 Excited to share that I’ve joined the amazing team at Skild AI! I’m already blown away by the energy, the quality of work, and the level of ambition here. The mission, vision, and results speak for themselves. This is just the beginning. More to come soon! 👇

ChatGPT for helping the Swedish Prime Minister:

📢Excited to share that I’ve joined @MBZUAI as an Assistant Professor of Computer Vision this fall! If you’re interested in CV4Science: building the next generation of foundation models & discovery tools for science, consider applying to MBZUAI. I’ll be recruiting PhD students!

#K2Think (🏔️💭) is now live. We're proud of this model that punches well above its weights, developed primarily for mathematical reasoning but has shown itself to be quite versatile. As a fully deployed reasoning system at k2think.ai you can test it for yourself!

“AI is evil” Meanwhile, ChatGPT:

imagine losing your job to this

Scalable GANs with Transformers arxiv.org/abs/2509.24935… hse1032.github.io/GAT authors train latent-space transformer GANs up to XL/2 scale, and report SotA 1-step class-conditional image generation results on ImageNet-256 after 40 epochs (*with REPA in discriminator)