Summary
MV-Forcing is a computer vision paper (arXiv:2607.05376) by Gal Fiebelman, Hadar Averbuch-Elor, and Sagie Benaim addressing long-range, multi-view consistent dynamic scene video generation. Existing video diffusion models either generate long monocular videos via temporal autoregression or short multi-view videos via bidirectional attention, but not both. MV-Forcing combines temporal and view autoregression within a single diffusion model by inserting a 4D geometric bridge between sequentially generated views. The key insight is that autoregressive 3D reconstruction models naturally interface between generated views: given a completed source view, the method reconstructs its 3D structure and renders a geometric prior for the next target view, which the diffusion model refines into high-quality video. The model is distilled via distribution matching distillation and spatio-temporal self-forcing, closing the train-inference exposure bias in both temporal and view-wise sequential autoregression. Experiments show the single-step student model generates geometrically consistent multi-view videos of dynamic scenes with arbitrary length and number of views.
Overview
- Field: Computer Vision
- Authors: Gal Fiebelman, Hadar Averbuch-Elor, Sagie Benaim
- Published: 2026-07-06
- arXiv: 2607.05376
Summary
Recent advances in video diffusion models have enabled either generating long monocular videos through temporal autoregression, or producing short multi-view videos through bidirectional attention. However, generating long-range, multi-view consistent videos of dynamic scenes remains an open problem.
This paper proposes MV-Forcing, which combines temporal and view autoregression within a single diffusion model by introducing a 4D geometric bridge between sequentially generated views. The key insight is that autoregressive 3D reconstruction models naturally serve as an interface between autoregressively generated views: given a completed source view, the method reconstructs its 3D structure and renders a geometric prior for the next target view, which the diffusion model refines into high-quality video.
The model is further distilled through distribution matching distillation and spatio-temporal self-forcing, closing the train-inference exposure bias that arises in both temporal and view-wise sequential autoregression.
Key Results
- MV-Forcing generates geometrically consistent multi-view videos of dynamic scenes.
- The single-step distilled student model supports arbitrary video length and an arbitrary number of views.
*Auto-collected on 2026-07-06.*
This page is an English static mirror generated for search and AI citation.
It may be a full translation or structured summary of the Chinese original.
Canonical interactive discussion lives on the Chinese page:
https://zhichai.net/topic/178346223