Paper Overview
Research Area: Computer Vision 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
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.
Key Contributions
- HDR (Hierarchical Denoising for Visual Reasoning): a unified framework that integrates hierarchical latents into causal video generation for multi-step reasoning.
- Video latents are organized into a tree-structured hierarchy, enabling coarse-to-fine reasoning before streaming output.
- Coarse denoising layers preserve uncertain hypotheses for global planning; finer layers progressively refine them into concrete visual states.
- A sparse hierarchical attention mechanism (SHAP) reduces temporal attention cost.
- Success rate: 34.22 → 60.29 vs. streaming autoregressive diffusion baseline (76.2% relative improvement), with average progress improving from 76.00 to 89.56 and more consistent reasoning trajectories.
- Low-latency streaming output maintained: 0.70 s per latent, with inference 54.2x faster than bidirectional diffusion.
- Data efficiency: 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 demonstrate HDR's potential for physical interaction and world modeling.
- arXiv: https://arxiv.org/abs/2607.15278
Benchmark and Results
The authors introduce a hierarchical multi-step video reasoning benchmark including out-of-distribution cases, covering six tasks: maze navigation, Tower of Hanoi, one-line drawing, sliding puzzle, Sokoban, and water pouring.