Paper Overview
Field: Computer Vision (CV) Authors: Zezhong Qian, Xiaowei Chi, Chak-Wing Mak, Tianze Zhou, Ruibin Yuan, Yuhan Rui, Hengzhe Sun, Zhuoqun Wu, Yuming Li, Siyuan Qian, Sirui Han, Shanghang Zhang Published: 2026-07-16 arXiv: 2607.15278
Summary (translated from Chinese)
Video models are evolving into vision foundation models, but they still lack human-like multi-step reasoning. Streaming autoregressive diffusion models are efficient yet limited in reasoning ability, while bidirectional diffusion enables global revision but is expensive at inference due to dense frame-level denoising. Both paradigms struggle to achieve logical consistency and low-latency streaming output for complex reasoning tasks.
This paper proposes HDR (Hierarchical Denoising for Visual Reasoning), a unified framework that integrates hierarchical latents into causal video generation for multi-step reasoning. HDR organizes video latents into a tree-structured hierarchy, enabling coarse-to-fine reasoning before streaming output. Coarse denoising layers preserve uncertain hypotheses for global planning, while fine layers progressively refine them into concrete visual states. A Sparse Hierarchical Attention mechanism (SHAP) further reduces temporal attention cost.
The authors also introduce a hierarchical multi-step video reasoning benchmark with out-of-distribution cases, covering six tasks: maze navigation, Tower of Hanoi, single-line drawing, sliding puzzle, Sokoban, and water pouring.
Key Results
- Success rate improves from 34.22 to 60.29 vs. streaming autoregressive diffusion baselines (76.2% relative improvement); average progress rises from 76.00 to 89.56, with more consistent reasoning trajectories.
- HDR maintains low-latency streaming output at 0.70 seconds per latent, 54.2x faster than bidirectional diffusion.
- With only 2% of training data, HDR retains 82.9% of full-data performance, while bidirectional diffusion retains only 52.0%.
- Real-world robot experiments further demonstrate HDR's potential in physical interaction and world modeling.
Original Abstract (excerpt)
> Video models are evolving into vision foundation models, yet they still lack human-like multi-step reasoning. Streaming autoregressive diffusion models are efficient but limited in reasoning, while bidirectional diffusion enables global revision with high inference costs due to dense frame-level denoising. Both paradigms struggle to achieve logical consistency and low-latency streaming for complex reasoning tasks. We propose HDR (Hierarchical Denoising for Visual Reasoning), a unified framework that integrates hierarchical latents into causal video generation for multi-step reasoning. HDR organizes video latents into a tree-structured hierarchy, enabling coarse-to-fine reasoning before streaming output. Coarse denoising layers preserve uncertain hypotheses for global planning, while finer ...
--- *Auto-collected on 2026-07-18*