Zun Wang

@ZunWang919

PhD student @unc @unccs @unc_ai_group; Previous @anucomputing, @ustc

Chapel Hill, NC

zunwang1.github.io

Replies Reposts Media

Pinned
Zun Wang
@ZunWang919
Feb 17
🚀 Excited to share AnchorWeave — a local-memory-augmented framework for world-consistent long-horizon video generation. - Global 3D reconstruction as memory accumulates cross-view misalignment and contaminates conditioning signals. - We replace a single noisy global 3D memory
00:00
1
Zun Wang
@ZunWang919
Jun 18
Thrilled that AnchorWeave is accepted to #ECCV2026!🎉 AnchorWeave tackles long-horizon memory for world-consistency modeling: instead of maintaining one noisy global 3D memory, it retrieves multiple local spatial 3D memories, avoiding multi-view geometric misalignment and
Zun Wang
@ZunWang919
Feb 17
🚀 Excited to share AnchorWeave — a local-memory-augmented framework for world-consistent long-horizon video generation. - Global 3D reconstruction as memory accumulates cross-view misalignment and contaminates conditioning signals. - We replace a single noisy global 3D memory
00:00
Zun Wang
@ZunWang919
Jun 1
Seeing ≠ knowing. 👀 Super fun project led by Yue — we built SpatialUncertain to test whether VLMs realize when a viewpoint is occluded or misleading, making a question unanswerable. Spoiler: they mostly don’t, and they don’t know why (i.e., they can’t pick a better view)
Yue Zhang
@zhan1624
Jun 1
🚨 Excited to share SpatialUncertain — a controlled framework for evaluating whether VLMs know when not to answer spatial questions (and why). ➡️ Spatial reasoning is not just about finding the right answer—it is about knowing whether the available evidence supports an answer at
Zun Wang
@ZunWang919
May 15
Check out PhyMotion, a physics-grounded reward for human video generation! RL is the new default for video generation, but it's ceilinged by the reward. VLM critics judge pixels but can miss physics (e.g. floating feet, self-penetration, impossible torques all slip through). We
00:00
Yidong Huang
@owenhuang117
May 15
🚨 Excited to introduce PhyMotion🤸: Structured 3D Motion Reward for Physics-Grounded Human Video Generation! ❌ Existing 2D video rewards misleadingly assign high scores to videos with floating feet, self-penetrating limbs, and physics-violating motions. ✅ PhyMotion lifts
Zun Wang
@ZunWang919
May 1
🎉 Excited to share EPiC is accepted to #ICML2026! We show that learning precise camera control for video diffusion doesn't need expensive 3D supervision or large-scale data. No camera or point cloud processing — just mask source videos based on visibility to construct precise
Zun Wang
@ZunWang919
May 29, 2025
🚨Thrilled to introduce EPiC🎥: Efficient Video Camera Control Learning with Precise Anchor-Video Guidance A generative model enables precise 3D camera trajectory control over user-provided videos or images. It achieves highly efficient training, completing within just 16
00:00
1