This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan.

MirroS launches Code-as-World framework for turning video into executable physics simulations
N
News EditorTechub News says AI research group MirroS has released Code-as-World, a framework that treats the physical world as an executable representation. The system converts video scenes into code that can run in the MuJoCo physics engine, producing a trio of object composition, physical evolution, and visual appearance. MirroS says the workflow uses an agent loop with hypothesis, instantiation, execution, rendering, and verification steps, running for up to five iterations to recover executable programs from real video. The verified world programs can then serve as training data with precise physics labels for vision-language models. The supervised Code-as-World-VL-9B model reached 55.4 MRA on the QuantiPhy validation set, topping Gemini 3.1 Flash’s 54.8. MirroS also said the code repository and model checkpoints are open sourced on GitHub.
Techub News said AI research group MirroS has released Code-as-World, a framework that treats the physical world as an executable representation.
The system turns video scenes into code that can run in the MuJoCo physics engine. MirroS describes the output as a trio made up of object composition, physical evolution, and visual appearance, aimed at representing physical mechanisms more precisely.
The framework uses an agent loop that goes through hypothesis, instantiation, execution, rendering, and verification. According to the release, it can recover executable programs from real video material in as many as five iterations.
MirroS said the verified world programs can be used as training data with precise physics labels for vision-language models. On that basis, the supervised Code-as-World-VL-9B model scored 55.4 MRA on the QuantiPhy validation set, above Gemini 3.1 Flash at 54.8.
The company also said the related code repository and model checkpoints are now open sourced on GitHub.
5400
Disclaimer:
The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.
Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.
