Prospective Forcing: Learning to Look Ahead for Autoregressive Long Video Diffusion

Yifan Shen; Yewei Xia; Gongxu Luo; Peiyuan Zhu; Boyang Sun; Zhenhao Chen; Guangyi Liu; Guangyi Chen

cs.CVOct 10, 2026

Prospective Forcing: Learning to Look Ahead for Autoregressive Long Video Diffusion

Authors: Yifan Shen, Yewei Xia, Gongxu Luo, Peiyuan Zhu, Boyang Sun, Zhenhao Chen, Guangyi Liu, Guangyi Chen,

Organizations: Mohamed bin Zayed University of AI · Institute of Foundation Models · Carnegie Mellon University

Abstract

Autoregressive video generators synthesize long videos through chunk-wise rollouts, yet they are typically trained to predict only the immediate future. This limited prediction horizon makes it challenging to maintain video quality over extended rollouts. To address this problem, we introduce Prospective Forcing, a multi-head prediction framework that injects long-range predictive signals into autoregressive video representations. Built on a rolling-window generator, Prospective Forcing attaches lightweight, offset-specific heads that share the same conditioning information and independently predict chunks at different future temporal offsets. By extending supervision beyond the next chunk, these heads encourage the shared history representation to retain information useful for longer-horizon prediction, supporting temporally consistent generation. At inference time, we accelerate sampling by using accepted future-chunk predictions to skip full-window backbone executions. We equip the head with learnable embeddings to generate multiple candidates for the next chunk, and a learned verifier either selects a candidate or rejects all and falls back to standard autoregressive generation.

Figures & tables

Figure 1: By learning to predict future, Prospective Forcing generates long videos that remain dynamic, temporally consistent, and visually stable from a single text prompt.

Figure 2: Temporal shortsightedness. The car gradually drives out of view with only next-chunk prediction, and foreseeing future chunks keeps it in frame.

Figure 3: Illustration of Prospective Forcing. Left : We jointly denoise $W=5$ chunks in window $r$ , yielding features $\mathbf{h}_{1}^{r}$ to $\mathbf{h}_{5}^{r}$ (green). Then $K=2$ heads predict $\hat{\mathbf{h}}_{6}^{r}$ and $\hat{\mathbf{h}}_{7}^{r}$ in parallel from $\mathbf{h}_{5}^{r}$ (orange). Leading clean chunks are committed and the window advances accordingly (blue). Right : $M=3$ candidates are generated for the next chunk $\hat{h}_{6}^{r}$ , and the verifier $V_{\phi}$ selects one or rejects all.

Figure 4: User study. Bars show the win rate of pairwise comparisons of human preference.

Figure 5: Qualitative comparison on a 30-second video generated from the prompt “A rally car taking a fast turn on a track”. Video comparison is provided in the supplementary materials.

Method	FPS $\uparrow$	VBench			VBench-Long	VideoScore	MUSIQ
Method	FPS $\uparrow$	Total $\uparrow$	Quality $\uparrow$	Semantic $\uparrow$	Quality $\uparrow$	Avg. $\uparrow$	Avg. $\uparrow$	Drift $\downarrow$
CausVid	9.28	76.22	77.02	73.00	77.43	2.637	66.49	9.02
Causal Forcing++	7.56	78.56	79.14	76.24	79.53	2.670	66.65	5.52
Self Forcing	7.79	78.27	78.76	76.33	78.24	2.905	68.84	6.18
LongLive	8.08	80.27	80.78	78.22	80.61	3.108	69.12	3.43
Rolling Forcing	7.95	78.63	80.05	72.94	82.20	2.775	68.92	5.32

Table 1: Quantitative comparison on 200 MovieGen prompts (30-second videos,

832\times 480

, 16 fps). Bold and underline denote the best and second-best results.

Horizons	Quality $\uparrow$	Drift $\downarrow$
$\{1\}$	81.00	4.08
$\{1,2\}$ (Ours)	81.29	3.29
$\{1,2,4\}$	80.21	2.89

Table 2: Ablation on prediction horizons.

Train	Inference	FPS $\uparrow$	Quality $\uparrow$	Drift $\downarrow$
Stage 1	$\ell=0$	7.69	81.29	3.29
	Unverified, $\ell=1$	12.03	81.80	4.22
	Unverified, $\ell=2$	14.51	81.16	4.51
	$\ell=0$	7.81	81.63	3.42
Stage 1 + 2	Unverified, $\ell=1$	11.95	81.98	4.14
	Verified, $\ell=1$ (Ours)	9.23	82.71	3.64

Table 3: Ablation on inference mode. Results of the model trained with stage 1 only and with both stages, under different inference modes.

Factor	Setting	Quality $\uparrow$	Subject $\uparrow$	Dynamic $\uparrow$	Imaging $\uparrow$	Drift $\downarrow$
Architecture	Shared recurrent head	77.14	84.72	100.0	63.47	5.10
Architecture	Sequential heads	78.37	85.30	100.0	65.01	4.16
Head input	3 chunks	79.58	85.48	100.0	65.74	4.04
Head input	5 chunks	77.25	84.41	100.0	64.13	4.61
Head output	3 chunks	80.43	94.23	17.0	71.40	2.58
Head output	5 chunks	80.87	95.28	23.5	70.88	2.05

Table 4: Ablation on design choices under stage

1

and standard inference. Each variant changes one component of the ODE-initialized model (row “ODE”).

Appendix figures & tables7 assets

Supplementary material from the paper’s appendix.

Appendix

Inference	Accept rate	Sampling passes	Cache-update passes	Speedup
Standard ( $\ell=0$ )	–	46	46	$1.00\times$
Unverified, $\ell=1$	100%	27	46	$1.56\times$
Verified (Ours)	32.1%	37	46	$1.20\times$

Appendix

Table 5: Per-video inference cost breakdown on a single A100 GPU (30-second video, 480 frames).

Stage 1 vs.	Visual Quality	Consistency	Dynamics	Avg.
Rolling Forcing	81.8%	86.4%	87.9%	85.4%
LongLive	59.6%	55.3%	68.1%	61.0%
Video-Mirai	78.8%	73.1%	86.5%	79.5%

Appendix

Table 6: User study of the stage-1 model with standard inference (

\ell=0

). Each entry is the percentage of pairwise comparisons in which participants preferred this model over the baseline. Average is computed over all comparisons of the three criteria.

Method	Subject $\uparrow$	Background $\uparrow$	Flickering $\uparrow$	Smoothness $\uparrow$	Dynamic $\uparrow$	Aesthetic $\uparrow$	Imaging $\uparrow$
CausVid	87.59	90.60	98.13	99.06	21.00	57.35	67.47
Causal Forcing++	85.33	90.48	95.93	97.71	72.00	60.33	66.13
Self Forcing	89.59	92.27	98.17	98.77	24.50	61.07	69.57
LongLive	89.70	92.97	96.20	97.89	62.00	62.95	69.34
Rolling Forcing	84.36	88.72	94.34	97.86	98.50	58.19	68.24
Video-Mirai	82.85	88.02	95.99	96.92	47.00	57.30	64.44

Appendix

Table 7: Per-dimension VBench quality scores (%).

Method	Object $\uparrow$	Multi-Obj. $\uparrow$	Action $\uparrow$	Spatial $\uparrow$	Scene $\uparrow$	App. Style $\uparrow$	Temp. Style $\uparrow$	Overall Cons. $\uparrow$
Valid videos	119	20	36	5	54	8	36	200
CausVid	92.12	77.50	94.44	77.28	35.07	21.29	22.19	23.51
Causal Forcing++	95.06	76.88	94.44	87.15	42.82	21.35	22.72	24.46
Self Forcing	95.48	65.31	91.67	90.20	48.73	21.17	24.17	24.85
LongLive	95.43	80.94	94.44	80.74	49.54	21.17	24.95	25.97
Rolling Forcing	87.61	53.44	91.67	78.49	51.62	21.34	24.06	25.05

Appendix

Table 8: Per-dimension VBench semantic scores (%). The first row reports the number of prompts to which each dimension applies.

Method	Subject $\uparrow$	Background $\uparrow$	Smoothness $\uparrow$	Dynamic $\uparrow$	Aesthetic $\uparrow$	Imaging $\uparrow$
CausVid	98.15	96.99	99.08	21.80	56.30	67.18
Causal Forcing++	96.09	95.76	97.74	60.77	58.86	65.39
Self Forcing	98.19	96.90	98.82	21.77	59.73	69.20
LongLive	97.20	96.26	97.96	55.50	61.43	68.66
Rolling Forcing	96.16	95.59	97.92	88.20	57.05	67.73
Video-Mirai	96.41	96.21	96.93	36.43	56.54	63.84

Appendix

Table 9: Per-dimension VBench-Long scores (%).

Method	Visual $\uparrow$	Temporal $\uparrow$	Dynamic $\uparrow$	T2V Align. $\uparrow$	Factual $\uparrow$
CausVid	2.683	2.450	2.985	2.738	2.330
Causal Forcing++	2.698	2.472	3.044	2.803	2.332
Self Forcing	2.908	2.731	3.242	3.046	2.597
LongLive	3.157	2.985	3.345	3.206	2.847
Rolling Forcing	2.843	2.608	3.095	2.876	2.452
Video-Mirai	2.679	2.461	3.187	2.752	2.441

Appendix

Table 10: Per-dimension VideoScore v1.1 scores.

Method	Head $\uparrow$	Tail $\uparrow$	Avg. $\uparrow$	$\frac{1}{N}\sum_{i}\|T_{i}-H_{i}\|\downarrow$
CausVid	70.25	62.73	66.49	9.02
Causal Forcing++	68.00	65.31	66.65	5.52
Self Forcing	71.06	66.61	68.84	6.18
LongLive	69.37	68.87	69.12	3.43
Rolling Forcing	69.48	68.35	68.92	5.32
Video-Mirai	69.49	59.58	64.54	11.90

Appendix

Table 11: MUSIQ scores of the first and last 5 seconds and the drift measure.

Autoregressive Video Generation

Autoregressive video generation creates videos frame by frame or in chunks, conditioning each prediction on prior context for streaming and interactive use.

See all 203 papers

Head Forcing: Long Autoregressive Video Generation via Head Heterogeneity
May 14, 2026Jiahao Tian, Yiwei Wang, Gang Yu +1Diffusion Transformer Autoregressive Diffusion
AGI Lab, Westlake University · University of California at Merced · StepFun
Connected Self Forcing: Beyond Local Learning in Video Autoregression
Oct 8, 2026Dongbin Zhang, Chaoda Zheng, Kangjie Chen +13Autoregressive Video Generation Streaming Video Generation
XPeng · The Chinese University of Hong Kong · Tsinghua University
Relax Forcing: Relaxed KV-Memory for Consistent Long Video Generation
Mar 22, 2026Zengqun Zhao, Yanzuo Lu, Ziquan Liu +3Long-Horizon Video Generation Autoregressive Video Generation
Queen Mary University of London · Imperial College London · Huawei R&D UK

May 14, 2026J/K move · Enter open · S save

Head Forcing: Long Autoregressive Video Generation via Head Heterogeneity

Jiahao Tian, Yiwei Wang, Gang Yu, Chi Zhang

AGI Lab, Westlake University · University of California at Merced · StepFun

Diffusion Transformer Autoregressive Diffusion Self-Attention Long-Horizon Video Generation KV-Cache Management Autoregressive Video Generation+2

Open paper AskPDF arXiv

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/.

Abstract

Autoregressive Video Generation

Explore similar work

Head Forcing: Long Autoregressive Video Generation via Head Heterogeneity

Connected Self Forcing: Beyond Local Learning in Video Autoregression

Relax Forcing: Relaxed KV-Memory for Consistent Long Video Generation