Super excited to share the last paper of my PhD: "Hallucination in World Models is Predictable and Preventable" ✨ We train a 350M-parameter generative world model on a large dataset spanning 210 tasks and show that we can predict *when* hallucination will happen and use that info to fix it! 🧵1/n

Interested in ✨world models✨? I just open-sourced an implementation of the Dreamer 4 world model. It's in PyTorch and comes with a pretrained model + a neat little web interface that lets you interact with any of 30 DMControl tasks that I trained it on! Link: github.com/nicklashanse...

🎉 New work: “Learning Massively Multitask World Models for Continuous Control” We introduce MMBench: a 200-task RL benchmark, and Newt: a language-conditioned multitask world model trained with large-scale online RL. www.nicklashansen.com/NewtWM/ Code, checkpoints, dataset etc. are open-source!

Introducing DINOv3 🦕🦕🦕 A SotA-enabling vision foundation model, trained with pure self-supervised learning (SSL) at scale. High quality dense features, combining unprecedented semantic and geometric scene understanding. Three reasons why this matters👇

Are you working on real-to-sim, sim-to-real, learning world models, or using physics based simulators to do interesting robotics tasks? There are 15 days to the submission deadline for our CoRL workshop Learning to Simulate Robot Worlds. More details here simulatingrobotworlds.github.io/submit.html

Learning to Simulate Robot Worlds

simulatingrobotworlds.github.io

🚀 We’re pleased to announce our workshop at CoRL 2025: Learning to Simulate Robot Worlds! Workshop website: simulatingrobotworlds.github.io The workshop aims to cover topics like physics-grounded simulation, photorealistic digital twins, AI-controlled simulators, to learned neural world models.

Learning to Simulate Robot Worlds

Join the Learning to Simulate Robot Worlds workshop.

simulatingrobotworlds.github.io

SCMs often assume overly-dense causal graphs in dynamic settings 👉⚽🥎, since any object interaction is a potential causal edge, making them hard to scale. In joint work with @turanorujlu.bsky.social @cgumbsch.bsky.social & Martin Butz we propose a new Causal Process Model to tackle this. Thread👇

Reframing attention as a reinforcement learning problem for causal discovery

Formal frameworks of causality have operated largely parallel to modern trends in deep reinforcement learning (RL). However, there has been a revival of interest in formally grounding the representati...

arxiv.org

Turan Orujlu@turanorujlu.bsky.social · last yr.

Reframing attention as an RL problem for causal discovery AI models like GNNs & Transformers can struggle with dynamic causal reasoning. Our work introduces the Causal Process Model (CPM), which reframes attention as an RL problem. Agents dynamically build sparse causal graphs.

Sergey Levine was just presenting in the Exploration in AI @ #ICML2025 and promoted that exploration needs to be grounded, and that VLMs are a good source ;-) Check our paper below 👇

Cansu Sancaktar@cansusancaktar.bsky.social · last yr.

✨Introducing SENSEI✨ We bring semantically meaningful exploration to model-based RL using VLMs. With intrinsic rewards for novel yet useful behaviors, SENSEI showcases strong exploration in MiniHack, Pokémon Red & Robodesk. Accepted at ICML 2025🎉 Joint work with @cgumbsch.bsky.social 🧵

I am going to present the poster during the next poster session. 11am Wed. Poster W #707

Cansu Sancaktar@cansusancaktar.bsky.social · last yr.

✨Introducing SENSEI✨ We bring semantically meaningful exploration to model-based RL using VLMs. With intrinsic rewards for novel yet useful behaviors, SENSEI showcases strong exploration in MiniHack, Pokémon Red & Robodesk. Accepted at ICML 2025🎉 Joint work with @cgumbsch.bsky.social 🧵

🚀 Launch day! The NeurIPS 2025 PokéAgent Challenge is live. @neuripsconf.bsky.social Two tracks: ① Showdown Battling – imperfect-info, turn-based strategy ② Pokemon Emerald Speedrunning – long horizon RPG planning 5 M labeled replays • starter kit • baselines. Bring your LLM, RL, or hybrid agent!

✨Introducing SENSEI✨ We bring semantically meaningful exploration to model-based RL using VLMs. With intrinsic rewards for novel yet useful behaviors, SENSEI showcases strong exploration in MiniHack, Pokémon Red & Robodesk. Accepted at ICML 2025🎉 Joint work with @cgumbsch.bsky.social 🧵

Mark your calendars, EWRL is coming to Tübingen! 📅 When? September 17-19, 2025. More news to come soon, stay tuned!