论文HF Daily Papers · arXiv 2611.262113 天前
SpecWM免训练推理框架实现流式世界模型1.63倍加速与导航评分提升
本文由 AI 根据公开内容整理,可能有误,请以原文为准:SpecWM: Turning Verification into Refinement for Streaming World Models(HF Daily Papers · arXiv 2611.26211)
核心内容
针对视频世界模型推测解码中外部评分维度受限及拒绝后重生成成本高昂的问题,研究团队提出Speculative World Model(SpecWM)免训练推理框架。该方法利用目标模型的潜在速度场直接验证草稿块,将验证过程转化为可复用的修正信号。在WBench Navigation基准测试中,SpecWM相比目标模型实现1.63倍推理加速,同时将导航评分(NavScore)由草稿模型的75.99显著提升至85.63,达成了质量、速度与控制性能的最优权衡。
技术上,SpecWM设置双阈值将草稿判定为接受、细化或拒绝,避免无效计算。一步细化复用修正信号提升感知质量,三步细化则增强相机跟随与轨迹一致性,并引入转换感知细化策略。LongSpecBench测试证实该框架在多轮长时序生成中保持稳定的质量控制能力;Hy-Nomas实验进一步验证了SpecWM在不同骨干网络架构及去噪预算下的通用性,为具身智能仿真与端到端规划提供了高效的世界模型推理范式。
关键看点
- 潜在速度场验证:利用目标模型Latent Velocity Field替代外部评分器,将验证结果转化为可复用修正信号,消除解码与打分额外开销
- 分级细化策略:通过双阈值机制执行Accept/Refine/Reject决策,一步细化优化感知质量,三步细化强化相机跟随与轨迹一致性
- 导航性能跃升:在WBench Navigation上NavScore从75.99升至85.63,推理速度达目标模型1.63倍,解决质量-速度-控制三角难题
- 跨架构泛化验证:LongSpecBench证实长时序多轮生成稳定性,Hy-Nomas实验确认框架适配不同Backbone与Denoising Budget
工程与产业观察
SpecWM通过挖掘目标模型内部动力学特征替代外部奖励反馈,解决了具身世界模型推理延迟与生成质量的矛盾。其免训练特性与潜在空间修正机制,为机器人端到端策略学习中的高保真、低延迟环境预测提供了关键工程解法,有助于加速Sim-to-Real闭环验证效率。
英文摘要 (Abstract)
Recently, video world models have advanced interactive generation, with large models offering higher quality and smaller models enabling faster inference. Existing speculative methods combine their strengths through fast draft generation and selective target regeneration guided by rewards on decoded frames. However, verification covers only the external scorer's evaluation dimensions and adds decoding and scoring costs. Moreover, these scores provide no correction, requiring costly target regeneration for rejected drafts. To address these limitations, we propose Speculative World Model (SpecWM), a unified, training-free inference framework for streaming world models. It verifies draft chunks using the target's latent velocity field. Verification also yields a reusable correction; two thresholds select Accept, Refine, or Reject. Specifically, one-step refinement reuses it to improve perceptual quality, while three-step refinement strengthens camera following and trajectory consistency. These findings motivate transition-aware refinement. On WBench Navigation, SpecWM improves visual quality at 1.63times speedup over the target and raises NavScore from the draft's 75.99 to 85.63, achieving the best quality-speed-control trade-off. Additionally, our LongSpecBench shows sustained quality and control over long, multi-turn rollouts, while experiments on Hy-Nomas validate SpecWM across backbones and denoising budgets.
Hugging Face Daily Papers 热度:1 upvotes 作者:Jiaming Li, Lingzhen Li, Xiaze Yang, Chenyu Zhu, Kaiqi Liu, Xiaotong Zhao, Songlian Li, Toyota Li, Alan Zhao, Haobin Zhong, Yue Ma, Zhiyuan Ma
外部来源
本页事实对应的公开页面,以原文为准。(点击将离开本站)