Head Forcing: Long Autoregressive Video Generation via Head Heterogeneity

• URL: http://arxiv.org/abs/2605.14487v1
• Date: Thu, 14 May 2026 07:27:39 GMT
• Title: Head Forcing: Long Autoregressive Video Generation via Head Heterogeneity
• Abstract summary: Autoregressive video diffusion models support real-time synthesis but suffer from error accumulation and context loss over long horizons.<n>We propose Head Forcing, a training-free framework that assigns each head type a tailored KV cache strategy.<n>Without additional training, Head Forcing extends generation from 5 seconds to minute-level duration, supports multi-prompt interactive synthesis, and consistently outperforms existing baselines.
• Score: 21.347969441374506
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Autoregressive video diffusion models support real-time synthesis but suffer from error accumulation and context loss over long horizons. We discover that attention heads in AR video diffusion transformers serve functionally distinct roles as local heads for detail refinement, anchor heads for structural stabilization, and memory heads for long-range context aggregation, yet existing methods treat them uniformly, leading to suboptimal KV cache allocation. We propose Head Forcing, a training-free framework that assigns each head type a tailored KV cache strategy: local and anchor heads retain only essential tokens, while memory heads employ a hierarchical memory system with dynamic episodic updates for long-range consistency. A head-wise RoPE re-encoding scheme further ensures positional encodings remain within the pretrained range. Without additional training, Head Forcing extends generation from 5 seconds to minute-level duration, supports multi-prompt interactive synthesis, and consistently outperforms existing baselines. Project Page: https://jiahaotian-sjtu.github.io/headforcing.github.io/.

Related papers

• FadeMem: Distance-Aware Memory Consolidation for Autoregressive Video Diffusion [59.207505503284715]
  FadeMem organizes historical KV blocks into a temporal hierarchy under a fixed cache budget.<n>New history is inserted as fine-grained entries, while older adjacent entries are progressively merged.<n>Experiments show improved subject consistency, background stability, and temporal coherence over existing bounded-cache strategies.
  arXiv  Detail & Related papers  (2026-06-09T10:22:18Z)
• DySink: Dynamic Frame Sinks for Autoregressive Long Video Generation [51.58559206569209]
  We propose DySink, a retrieval-based framework that maintains a compact memory bank and selects visually relevant historical frames as dynamic frame sinks.<n> Experiments on minute-long videos show that DySink consistently improves dynamic degree over strong baselines while also achieving higher temporal quality.
  arXiv  Detail & Related papers  (2026-05-20T11:01:01Z)
• Attend Locally, Remember Linearly: Linear Attention as Cross-Frame Memory for Autoregressive Video Diffusion [61.57938553036056]
  ARL2 is a hybrid attention module that replaces quadratic cross-frame attention with a fixed-size recurrent state.<n>We decompose self-attention into two branches: an intra-frame softmax branch for spatial detail and local dependencies, and an inter-frame recurrent linear branch that maintains a fixed-size state for streaming context.<n>With 75% of layers replaced by hybrid linear attention, the model achieves up to 2.26 wall-clock speedup and 54% memory reduction, while maintaining comparable quality with improving temporal consistency.
  arXiv  Detail & Related papers  (2026-05-15T19:33:45Z)
• Pyramid Forcing: Head-Aware Pyramid KV Cache Policy for High-Quality Long Video Generation [14.413680197991356]
  Pyramid Forcing is a head-aware KVCache framework that identifies head types offline and assigns behavior-specific cache policies.<n> Experiments show that Pyramid Forcing consistently improves long-horizon generation quality on VBench-Long.
  arXiv  Detail & Related papers  (2026-05-13T07:23:02Z)
• LoGeR: Long-Context Geometric Reconstruction with Hybrid Memory [97.14005794889134]
  We present LoGeR, a novel architecture that scales dense 3D reconstruction to extremely long sequences without post-optimization.<n>LoGeR processes video streams in chunks, leveraging strong bidirectional priors for high-fidelity intra-chunk reasoning.<n>This memory architecture enables LoGeR to be trained on sequences of 128 frames, and generalize up to thousands of frames during inference.
  arXiv  Detail & Related papers  (2026-03-03T18:55:37Z)
• Towards Holistic Modeling for Video Frame Interpolation with Auto-regressive Diffusion Transformers [95.68243351895107]
  We propose a holistic, video-centric paradigm named textbfLocal textbfDiffusion textbfForcing for textbfVideo textbfFrame textbfInterpolation (LDF-VFI)<n>Our framework is built upon an auto-regressive diffusion transformer that models the entire video sequence to ensure long-range temporal coherence.<n>LDF-VFI achieves state-of-the-art performance on challenging long-sequence benchmarks, demonstrating superior per
  arXiv  Detail & Related papers  (2026-01-21T12:58:52Z)
• HeteroCache: A Dynamic Retrieval Approach to Heterogeneous KV Cache Compression for Long-Context LLM Inference [14.17979669446161]
  We propose HeteroCache, a training-free dynamic compression framework.<n>We show that HeteroCache achieves state-of-the-art performance on multiple long-context benchmarks and accelerates decoding by up to $3times$ compared to the original model in the 224K context.
  arXiv  Detail & Related papers  (2026-01-20T07:35:06Z)
• Lethe: Layer- and Time-Adaptive KV Cache Pruning for Reasoning-Intensive LLM Serving [11.750209684686707]
  Generative reasoning with large language models (LLMs) often involves long decoding sequences.<n>We propose Lethe, a dynamic KV cache management framework that introduces adaptivity along both the spatial and temporal dimensions of decoding.<n>Lethe achieves a favorable balance between efficiency and generation quality across diverse models and tasks, increases throughput by up to 2.56x.
  arXiv  Detail & Related papers  (2025-11-08T14:52:43Z)
• LoViC: Efficient Long Video Generation with Context Compression [68.22069741704158]
  We introduce LoViC, a DiT-based framework trained on million-scale open-domain videos.<n>At the core of our approach is FlexFormer, an expressive autoencoder that jointly compresses video and text into unified latent representations.
  arXiv  Detail & Related papers  (2025-07-17T09:46:43Z)
• Beyond Homogeneous Attention: Memory-Efficient LLMs via Fourier-Approximated KV Cache [67.47789629197857]
  We propose a training-free framework that exploits the heterogeneous roles of transformer head dimensions.<n>By projecting the long-context-insensitive dimensions onto Fourier bases, FourierAttention approximates their temporal evolution with fixed-length spectral coefficients.<n>We show that FourierAttention achieves the best long-context accuracy on LongBench and Needle-In-A-Haystack.
  arXiv  Detail & Related papers  (2025-06-13T15:35:54Z)
• Exploiting Temporal State Space Sharing for Video Semantic Segmentation [53.8810901249897]
  Video semantic segmentation (VSS) plays a vital role in understanding the temporal evolution of scenes.<n>Traditional methods often segment videos frame-by-frame or in a short temporal window, leading to limited temporal context, redundant computations, and heavy memory requirements.<n>We introduce a Temporal Video State Space Sharing architecture to leverage Mamba state space models for temporal feature sharing.<n>Our model features a selective gating mechanism that efficiently propagates relevant information across video frames, eliminating the need for a memory-heavy feature pool.
  arXiv  Detail & Related papers  (2025-03-26T01:47:42Z)
• Model Tells You What to Discard: Adaptive KV Cache Compression for LLMs [82.08922896531618]
  We introduce adaptive KV cache compression, a plug-and-play method that reduces the memory footprint of generative inference for Large Language Models (LLMs) We conduct targeted profiling to discern the intrinsic structure of attention modules. Based on the recognized structure, we then construct the KV cache in an adaptive manner: evicting long-range contexts on attention heads emphasizing local contexts, discarding non-special tokens on attention heads centered on special tokens, and only employing the standard KV cache for attention heads that broadly attend to all tokens.
  arXiv  Detail & Related papers  (2023-10-03T05:17:08Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.