跳到正文
HuggingFace Daily Papers(社區熱門論文)·· 2026-10-15AI 評分46

Speculative Forcing for Efficient Autoregressive Video Generation

Speculative Forcing for Efficient Autoregressive Video Generation

AI 導讀

Speculative Forcing 透過提早接受部分 token 草稿、跳過不必要的去噪步驟,加速少步自回歸影片生成,同時維持完整步數採樣的品質。LongLive 每個 token 約需 1.7 個去噪步驟,速度提升超過 1.5 倍,達 41.7 和 37.8 fps。Lingbot-World 2.0 每個 token 約需 2.5 個去噪步驟,速度提升 1.35 倍,達 24.5 fps。

正文

Abstract

Speculative Forcing accelerates few-step autoregressive video generation by accepting early token drafts and skipping unnecessary denoising steps while preserving output quality.

Few-step autoregressive video generators enable streaming generation, yet they run every token through the final denoising step, regardless of how much refinement it needs. We find this uniform computation is often unnecessary: many tokens already closely match their eventual full-step outputs in the initial steps, while others require refinement until the end. We introduce Speculative Forcing, a framework that treats intermediate clean estimates as drafts and predicts when individual tokens can be safely accepted early. Once accepted, a token is frozen as its final output and undergoes no further denoising, while its cached keys and values remain available as context for the remaining tokens. We explore two alternative verifier designs: a training-free one based on similarity to clean content and stability across steps, and a learned one that reads the frozen generator's intermediate features to estimate each token's remaining gap to its full-step output. Across short-video, long-video, and world generation, Speculative Forcing preserves the quality of full-step sampling while substantially reducing computation. On LongLive, it matches the four-step generator with ~1.7 denoising steps per token, running over 1.5x faster at 41.7 and 37.8 fps. On Lingbot-World 2.0, where camera motion continually changes the visible content, it maintains quality with ~2.5 denoising steps per token and runs 1.35x faster at 24.5 fps.

View arXiv page View PDF Project page GitHub 2 Add to collection

Get this paper in your agent:

hf papers read 2610.22347

Don't have the latest CLI?

curl -LsSf https://hf.co/cli/install.sh | bash

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2610.22347 in a model README.md to link it from this page.

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2610.22347 in a dataset README.md to link it from this page.

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2610.22347 in a Space README.md to link it from this page.

Collections including this paper 1

來源:HuggingFace Daily Papers(社區熱門論文) · huggingface.co