RAYNOVA: Scale-Temporal Autoregressive World Modeling in Ray Space

CVPR 26 2026 65.8 method, application

TLDR

RAYNOVA proposes a dual-causal autoregressive world model in ray space for multi-view driving video generation, achieving state-of-the-art results on nuScenes.

Reasoning

The paper introduces a novel geometry-agnostic representation using Plücker-ray encoding and a recurrent training paradigm to handle long-horizon generation. Strengths include strong generalization to novel camera setups and high throughput; weaknesses are limited to driving scenarios and lack of interactive or RL-based evaluation.

Read-first score

Read-first score 65.8, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 42.

Recency 8%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Methodology quality 25%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=dataset,evaluation,metric,result

Topical relevance 42%

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Reproducibility 25%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=code,dataset,github

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 237.

Keyword Scores

world model

video world model

generative world model

world dynamics prediction

world simulator

interactive world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Dual-causal autoregressive framework that follows both scale-wise and temporal topological orders for unified 4D spatio-temporal reasoning
Isotropic spatio-temporal representation across views, frames, and scales using relative Plücker-ray positional encoding, enabling geometry-agnostic generalization to diverse camera setups and ego motions
Recurrent training paradigm to alleviate distribution drift in long-horizon video generation

Methodology

RAYNOVA is a geometry-agonistic multiview world model for driving scenarios. It employs a dual-causal autoregressive framework with global attention, processing spatial and temporal correlations jointly via scale-wise and temporal topological orders. The model uses relative Plücker-ray positional encoding to construct an isotropic spatio-temporal representation, and a recurrent training paradigm is introduced to mitigate distribution drift. Evaluation is performed on the nuScenes dataset for multi-view video generation, with comparisons to prior methods.

Key Results

RAYNOVA achieves state-of-the-art multi-view video generation results on nuScenes, while offering higher throughput and strong controllability under diverse input conditions, and generalizing to novel views and camera configurations without explicit 3D scene representation.

Resources

Paper Project