Lip Forcing distills a 14B bidirectional video diffusion teacher into autoregressive students that achieve real-time lip synchronization at 31 FPS using two denoising steps without CFG.
hub Canonical reference
Rolling Forcing: Autoregressive Long Video Diffusion in Real Time
Canonical reference. 83% of citing Pith papers cite this work as background.
60
Pith papers citing it
Background
83% of classified citations
abstract
Streaming video generation, as one fundamental component in interactive world models and neural game engines, aims to generate high-quality, low-latency, and temporally coherent long video streams. However, most existing work suffers from severe error accumulation that often significantly degrades the generated stream videos over long horizons. We design Rolling Forcing, a novel video generation technique that enables streaming long videos with minimal error accumulation. Rolling Forcing comes with three novel designs. First, instead of iteratively sampling individual frames, which accelerates error propagation, we design a joint denoising scheme that simultaneously denoises multiple frames with progressively increasing noise levels. This design relaxes the strict causality across adjacent frames, effectively suppressing error growth. Second, we introduce the attention sink mechanism into the long-horizon stream video generation task, which allows the model to keep key value states of initial frames as a global context anchor and thereby enhances long-term global consistency. Third, we design an efficient training algorithm that enables few-step distillation over largely extended denoising windows. This algorithm operates on non-overlapping windows and mitigates exposure bias conditioned on self-generated histories. Extensive experiments show that Rolling Forcing enables real-time streaming generation of multi-minute videos on a single GPU, with substantially reduced error accumulation.
hub tools
citation-role summary
background 11
extension 1
citation-polarity summary
representative citing papers
VideoMLA applies multi-head latent attention with 3D-RoPE decoupling to autoregressive video diffusion, delivering 92.7% KV memory reduction while matching short-horizon baselines and leading long-horizon VBench scores.
ISPA reduces KV cache size by up to 50% in AR video models by transitioning layers to local attention and applying instance-specific least-squares weight modulation to compensate for lost history.
MemLearner introduces a learning-based adaptive context query method using query tokens in video world models to improve long-term scene consistency over rule-based retrieval.
TetherCache organizes KV-cache into sink, memory, and recent regions and applies gated recall with attention-diversity balancing plus trusted memory editing to stabilize long-horizon autoregressive video diffusion.
FadeMem introduces distance-aware KV memory consolidation for autoregressive video diffusion that builds a temporal hierarchy with power-law merging to preserve short-term dynamics and long-range coherence under fixed cache budget.
SwiftVR achieves real-time generative video restoration at 1080p on consumer GPUs (26 FPS on RTX 5090) and higher resolutions on H100 via efficient dense attention and chunk-wise autoencoding.
SPAWN enables training-free insertion of custom visual concepts into autoregressive world models by swapping the pinned context-memory anchor over a short injection window.
LongLive-RAG formulates long video generation as retrieval-augmented generation by treating self-generated latents as a dynamic searchable history and adding a Window Temporal Delta Loss for better retrieval.
AdaState replaces the static first-frame KV anchor with an evolving hidden latent that the model denoises alongside content, treating time as relative to enable recurrence and richer dynamics in streaming video generation.
Future Forcing constructs a future query proxy from historical pre-RoPE statistics to score and merge KV tokens, improving subject consistency by up to 1.49 on VBench-Long for 60s AR video generation.
Q-ARVD introduces final-quality-aware frame weighting and outlier-aware adaptive dual-scale quantization to enable accurate low-bit inference for autoregressive video diffusion models.
DySink maintains a memory bank and retrieves relevant historical frames as dynamic sinks while using an anomaly gate to suppress collapse, yielding higher temporal quality and dynamic degree on minute-long videos.
LongLive-2.0 delivers an NVFP4 parallel infrastructure that enables direct training of long multi-shot autoregressive diffusion video models and achieves up to 2.15x training and 1.84x inference speedups on Blackwell and other GPUs.
StreamingEffect enables real-time 720p human-centric video effect generation on one GPU via teacher-student distillation, keyframe control, and a new 130K video dataset.
Echo-Forcing decouples stable anchors, compressed history, and recent dynamics in video diffusion KV caches using hierarchical memory, scene recall frames, and difference-aware decay to support interactive long video generation under bounded cache.
EntityBench is a new benchmark with detailed per-shot entity schedules from real media, and the EntityMem baseline using persistent per-entity memory achieves the highest character fidelity with Cohen's d of +2.33.
CausalCine enables real-time causal autoregressive multi-shot video generation via multi-shot training, content-aware memory routing for coherence, and distillation to few-step inference.
FreeSpec uses SVD-based spectral reconstruction to fuse global low-rank and local high-rank features, reducing content drift and preserving temporal dynamics in long video generation.
Stream-R1 improves distillation of autoregressive streaming video diffusion models by adaptively weighting supervision with a reward model at both rollout and per-pixel levels.
A survey that groups efficient video diffusion methods into four paradigms—step distillation, efficient attention, model compression, and cache/trajectory optimization—and outlines open challenges for practical use.
Grounded Forcing introduces dual memory caching, reference-based positional embeddings, and proximity-weighted recaching to bridge stable semantics with local dynamics, improving long-range consistency in autoregressive video synthesis.
DreamDojo is a foundation world model pretrained on the largest human video dataset to date that uses continuous latent actions to transfer interaction knowledge and achieves controllable physics simulation after robot post-training.
ABot-M0.5 proposes a unified mobility-and-manipulation world action model using three alignment strategies that achieves state-of-the-art performance on mobile and fine-grained manipulation benchmarks.
citing papers explorer
-
DreamDojo: A Generalist Robot World Model from Large-Scale Human Videos
DreamDojo is a foundation world model pretrained on the largest human video dataset to date that uses continuous latent actions to transfer interaction knowledge and achieves controllable physics simulation after robot post-training.
-
AR Forcing: Towards Long-Horizon Robot Navigation World Model
AR Forcing trains diffusion world models by integrating standard noise prediction loss into an autoregressive loop that uses self-generated predictions as context, reducing train-inference mismatch for improved long-horizon image consistency and trajectory accuracy on navigation datasets.