EonWeave: Weaving Persistent Evidence into Consistent Long-Horizon World Generation

Long-horizon world generation requires a persistent model of the world, not merely short-range temporal continuity: a model must faithfully recover previously observed identities, scenes, and world states after they have disappeared from view for extended periods. We term this capability $\text{revisitation consistency}$. Existing autoregressive generators struggle in this regime because short-term conditions progressively accumulate errors, while historical evidence is either underutilized, redundantly stored, or inefficiently retrieved. To address these coupled limitations, we introduce $\text{EonWeave}$, a unified $\text{trust–preserve–align}$ framework for integrating persistent evidence into long-horizon generation. $\text{ARiSE}$ first calibrates source reliability by selectively re-noising the short-term condition while preserving clean historical evidence, preventing memory bypass and promoting adaptive reliance on persistent context. Building on this reliability asymmetry, $\text{ReCAP}$ curates a fixed-capacity Persistent Scene Atlas through pose–appearance diversity-gain selection and block-adaptive projection, retaining complementary world-state evidence under bounded memory. $\text{RaDAR}$ then converts the curated history into query-aligned references through noise-adaptive appearance–geometry routing and retrieval-to-reference distillation, enabling efficient utilization of heterogeneous historical evidence. Extensive experiments demonstrate state-of-the-art generation quality and revisitation consistency, with $\text{EonWeave}$'s PSNR advantage widening from approximately $3$ dB in single-shot generation to $6$ dB under extended autoregressive rollouts. These results establish $\text{EonWeave}$ as a substantive advance toward consistent long-horizon world modeling.