Video World Models · 2026-10-26
$\Psi$-0.6: Adaptive Streams of Thought in a Unified Vision-Language World Model
Khai Loong Aw, Klemen Kotar, Baihan Zhang, Yifan Liu, Lilian Chen, Atlas Kazemian, Rahul Venkatesh, Daniel Yamins
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
Ψ-0.6 is a unified vision-language world model that adaptively composes perception, language, geometry, and counterfactual simulation in a single shared token interface using streams of thought. Using reinforcement learning on task feedback, it learns to select and use intermediate geometric and imagined representations, improving zero-shot multi-view spatial reasoning and counterfactual visual world modeling results.
Why it matters
The authors target a limitation of fragmented physical AI systems by unifying perception, VLM-style language interaction, geometric representations, and simulation in one model. With an any-to-any prompting interface, intermediate predictions (e.g., camera pose, depth, optical flow, point tracks, and imagined views) become composable reasoning steps. Reinforcement learning then optimizes which intermediate representations to generate and how to use them, enabling adaptive reasoning strategies for downstream tasks.
Method
- Unified model: images, camera pose, depth, optical flow, point tracks, and text are serialized as tokens and processed by a single ~8B-parameter autoregressive transformer.
- Streams of thought: a generic any-to-any prompting interface specifies provided context and requested outputs; the output of one conditional prediction can be used as context for the next.
- Adaptive reasoning via RL: the policy selects and orders intermediate visual representations (depth, camera motion, optical flow) from task feedback to improve answer quality.
Abstract (from arXiv)
Humans reason about the physical world by flexibly combining different forms of inference as a problem unfolds. Achieving this flexibility is a major open challenge in AI: capabilities often remain separated across specialized models, limiting how they can be composed and jointly adapted to a task. We introduce $\Psi$-0.6, a unified vision-language world model that adaptively composes perception, language, geometry, and simulation within a shared reasoning process we call streams of thought. These streams interleave images, text, and explicit representations of geometry and motion, all of which the model can generate and interpret through a shared token interface. $\Psi$-0.6 surpasses leading open-weight models zero-shot on multi-view spatial reasoning, with further gains from reasoning through geometric intermediates. It also supports counterfactual simulation: generating and interpreting imagined scenes to determine, for example, whether moving an object would obstruct a pathway. Reinforcement learning from task feedback jointly optimizes the selection, generation, and use of intermediate representations, improving visual question-answering accuracy with gains that transfer to unseen benchmarks. $\Psi$-0.6 offers a path toward adaptive physical reasoning -- not only predicting the physical world but also choosing the representations through which to reason.
Related papers
- CoWorld Reasoner: Agentic Multimodal Reasoning with the Evolving Video World Model-Driven Harness
- WOVEN: Weaving Visual World Modeling into Multimodal LLMs
- Asking the World: Generalist Physical Reasoning through Agentic World Modeling and Probing
- Representing the Unseen: Rethinking Visual Imagination for Spatial Reasoning
- Vision-Language Agents Can Learn in Their Own Latent Worlds
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: Date not supplied · Indexed by Pub-AI: 2026-10-26 UTC
$Ψ$-0.6: Adaptive Streams of Thought in a Unified Vision-Language World Model
Source date: Date not supplied · Original record ↗
- arxiv
- 2610.32177
Imported from the existing registry. No new publication milestone asserted.
Abstract · arxiv
Humans reason about the physical world by flexibly combining different forms of inference as a problem unfolds. Achieving this flexibility is a major open challenge in AI: capabilities often remain separated across specialized models, limiting how they can be composed and jointly adapted to a task. We introduce $Ψ$-0.6, a unified vision-language world model that adaptively composes perception, language, geometry, and simulation within a shared reasoning process we call streams of thought. These streams interleave images, text, and explicit representations of geometry and motion, all of which the model can generate and interpret through a shared token interface. $Ψ$-0.6 surpasses leading open-weight models zero-shot on multi-view spatial reasoning, with further gains from reasoning through geometric intermediates. It also supports counterfactual simulation: generating and interpreting imagined scenes to determine, for example, whether moving an object would obstruct a pathway. Reinforcement learning from task feedback jointly optimizes the selection, generation, and use of intermediate representations, improving visual question-answering accuracy with gains that transfer to unseen benchmarks. $Ψ$-0.6 offers a path toward adaptive physical reasoning -- not only predicting the physical world but also choosing the representations through which to reason.