arXiv:2609.31654v2 Announce Type: replace-cross Abstract: Video diffusion transformers depend on temporal attention to coordinate information across frames, yet nearly everything known about this mec...
arXiv:2609.23658v1 Announce Type: cross Abstract: Despite impressive visual quality, state-of-the-art video diffusion models often generate content that violates real-world physical laws. While exist...
World in World introduces a training‑free inference interface that lets users control autoregressive video world models from new viewpoints. By converting diverse control signals—source‑video observations, target‑view projections, geometry renderings, and retrieved states—into camera‑ and time‑labelled visual tokens, the system uses a frozen causal video model’s self‑attention to maintain synchronization, complete unseen regions, and recover past appearances. The method supports camera‑controlled rerendering, long‑horizon revisiting, and human‑motion transfer while preserving perceptual quality, temporal consistency, and camera‑following accuracy.
arXiv:2609.37925v1 Announce Type: cross Abstract: Autoregressive (AR) video diffusion enables low-latency, streamable video generation, but prediction errors often accumulate over long rollouts. Trai...
arXiv:2403. 07711v5 Announce Type: replace-cross Abstract: Given the remarkable achievements in image generation through diffusion models, the research community has shown increasing interest in extending these models to video generation.
arXiv:2608.20473v1 Announce Type: new Abstract: Video language models process videos as dense visual-token sequences with substantial representational redundancy. Compressing these sequences is there...
The paper introduces Relax Forcing, a training‑free memory mechanism for autoregressive video diffusion that structures temporal context into Sink, Tail, and History frames. By selecting History frames with a relaxation criterion, the method reduces error accumulation and attention overhead while preserving motion dynamics. Experiments on VBench‑Long demonstrate that this structured memory improves long‑video generation quality over existing baselines.
World in World introduces a training‑free, inference‑time interface that transforms diverse control signals—such as source‑video observations, target‑view projections, geometry renderings, and retrieved states—into camera‑ and time‑labelled visual states. These states are processed by a frozen causal video model’s self‑attention, enabling tasks like camera‑controlled rerendering, long‑horizon revisiting, and human‑motion transfer without additional training. The method employs a correspondence router and evidence‑wise attention to align token identities and regulate auxiliary channel contributions during a single denoising pass.
arXiv:2605.25333v3 Announce Type: replace Abstract: Video world models should maintain evolving states when evidence is unobserved, yet current generators often freeze hidden states upon interruption...
RECAP-Forcing is a new method for long autoregressive video generation that addresses the memory challenge by organizing memory based on appearance novelty rather than recency. The approach retains key-value caches for newly appearing content—such as entering subjects, disoccluded regions, and new scenes—at the moment they first appear, ensuring consistent identities over time. It combines an attention sink for the initial scene with an optical-flow-based novelty bank for later frames, improving visual quality and semantic fidelity without adding learnable parameters.
StarWM introduces a self‑supervised attention routing mechanism that selectively applies reconstruction only to dynamically relevant regions of visual input. By combining a cross‑attention module with a dual‑stream decoder and stop‑gradient barriers, it balances faithful environmental dynamics capture with abstraction of irrelevant content. Experiments on DeepMind Control show that StarWM outperforms both reconstruction‑based and reconstruction‑free baselines, especially under distractor conditions, and preserves state attributes over long‑horizon imagination.
The paper introduces REVEAL, a diagnostic benchmark that stresses Video‑Language Models (VidLMs) on five controlled probes—camera‑motion sensitivity, cross‑frame integration, video sycophancy, language‑only shortcuts, and temporal expectation bias—to assess how well these models encode and use visual evidence. Experiments on 12 VidLMs reveal systematic failures: some visual signals are never reliably encoded, while others are overridden by model priors, leading to performance below chance on several probes that humans solve with high accuracy. Mechanistic probes further pinpoint where and why visual evidence is lost, demonstrating that under assertive prompts a model’s output becomes nearly invariant to real versus random video input, rendering visual evidence causally inert.