A new arXiv preprint introduces AEON, a framework that aims to unify video and 3D world models for explicit world simulation. According to the abstract, current approaches follow two distinct pathways: video generative models such as Sora, which simulate realistic appearances but do not maintain 3D consistency, and 3D generative models, which take the opposite approach.
The abstract is truncated, so the full details of AEON's method are not available in the source. However, the title and framing make clear that the paper's central contribution is to combine these two pathways into a single model, rather than treating them as separate tracks.
The motivation is straightforward: explicit world simulation needs both visual realism and geometric coherence. By unifying video and 3D world modeling, AEON could address a known limitation of each approach, though the paper's specific techniques and results are not described in the available text.