arXiv:2602. 01801v2 Announce Type: replace-cross Abstract: Autoregressive video diffusion models enable streaming generation, opening the door to long-form synthesis, video world models, and interactive neural game engines.
arXiv:2610.14654v1 Announce Type: cross Abstract: Autoregressive video generators synthesize long videos through chunk-wise rollouts, yet they are typically trained to predict only the immediate futu...
arXiv:2608. 08469v1 Announce Type: new Abstract: Existing streaming multimodal models process observations incrementally but still follow a turn-based prefill-then-decode pattern, making them non-duplex: new observations cannot naturally enter an active generation stream.
arXiv:2610.20137v1 Announce Type: new Abstract: Current video object removal methods suffer from prohibitive latency (minutes per video) and non-causal processing that prevents real-time interaction....
arXiv:2606. 16353v1 Announce Type: cross Abstract: Streaming video understanding models must answer queries at any moment during an ongoing stream, using only what they have observed so far and under fixed memory and computation budgets.
arXiv:2610.28323v1 Announce Type: new Abstract: Streaming video understanding requires a model to decide not only what to answer but also when to answer. Evidence arrives progressively, and recent be...
Current video object removal methods suffer from prohibitive latency (minutes per video) and non-causal processing that prevents real-time interaction. We present StreamErase, enabling sub-second late...
The paper introduces DVG, a Dynamic Video Generation framework that allocates computation across time and space to accelerate diffusion-based video generation. By selecting content-aware acceleration strategies automatically, DVG achieves near-lossless speedups—up to 7× on HunyuanVideo models and 18× when combined with distillation—without manual tuning or retraining. The authors provide code on GitHub for reproducibility.
The paper introduces SPOT‑Bench, a new benchmark for streaming video that evaluates models on proactive, multi‑turn queries rather than retrospective, fixed‑timestamp questions. It proposes the Timeliness‑F1 metric to assess temporal precision and coverage across the entire video. The study finds that current multimodal large language models (MLLMs) can detect events reliably but often spam predictions, and that post‑training silence reduces spamming at the cost of responsiveness, revealing a significant dead‑time issue in streaming inference.
arXiv:2607. 20174v1 Announce Type: cross Abstract: Existing human--object interaction (HOI) video generation methods are largely limited to offline short-video generation with complex driving conditions, making them unsuitable for real-time interactive applications.
arXiv:2606. 26762v1 Announce Type: cross Abstract: Streaming video understanding (SVU) must answer queries that arrive asynchronously while visual tokens stream continuously under strict GPU-memory and query-time latency budgets.
arXiv:2609.37559v1 Announce Type: new Abstract: To serve as real-world personal assistants, streaming video models need persistent memory that retains past experiences for later use. Yet existing str...