UniClawBench: A Universal Benchmark for Proactive Agents on Real-World Tasks \ Preprint, 2026 \ Zhekai Chen^*, Chengqi Duan^*, Kaiyue Sun^*, Bohao Li, Yuqing Wang, Manyuan Zhang^†, Xihui Liu^†

project | arXiv | github

A capability-driven benchmark containing 400 bilingual, real-world tasks for proactive agents. Evaluates skill usage, exploration, long-context reasoning, multimodal understanding, and cross-platform coordination.

MACRO: Advancing Multi-Reference Image Generation with Structured Long-Context Data \ Preprint, 2026 \ Zhekai Chen, Yuqing Wang, Manyuan Zhang^†, Xihui Liu^†

project | arXiv | github

Introduces MacroData, 400K structured samples with as many as ten reference images. Covers customization, illustration, spatial reasoning, and temporal prediction under long multimodal context.

Wan-Move: Motion-controllable Video Generation via Latent Trajectory Guidance \ NeurIPS 2025 \ Ruihang Chu^*†‡, Yefei He^*, Zhekai Chen^*, Shiwei Zhang^†, Xiaogang Xu, Bin Xia, Dingdong Wang, Hongwei Yi, Xihui Liu, Hengshuang Zhao, Yu Liu, Yingya Zhang, Yujiu Yang^†

project | arXiv | github

Enables precise object and camera motion control using latent trajectory guidance. Integrates motion-aware conditions into an off-the-shelf image-to-video model without architectural changes.

TTS-VAR: A Test-Time Scaling Framework for Visual Auto-Regressive Generation \ NeurIPS 2025 \ Zhekai Chen, Ruihang Chu^†, Yukang Chen, Shiwei Zhang, Yujie Wei, Yingya Zhang, Xihui Liu^†

arXiv | github

The first general test-time scaling framework for visual autoregressive models. Formulates generation as path search with diversity exploration and potential-based selection.

Framer: Interactive Frame Interpolation \ ICLR 2025 \ Wen Wang, Qiuyu Wang, Kecheng Zheng, Hao Ouyang, Zhekai Chen, Biao Gong, Hao Chen, Yujun Shen, Chunhua Shen

project | arXiv | github

Interactive frame interpolation with flexible drag-based local motion control. Produces diverse transitions from the same start and end frames and supports image morphing.

FreeCompose: Generic Zero-Shot Image Composition with Diffusion Prior \ ECCV 2024 \ Zhekai Chen^*, Wen Wang^*, Zhen Yang, Zeqing Yuan, Hao Chen^†, Chunhua Shen^†

arXiv | github

A generic zero-shot framework for appearance and semantic image composition using diffusion priors. Extends to object removal and multi-character customization without task-specific training.

AutoStory: Generating Diverse Storytelling Images with Minimal Human Effort \ International Journal of Computer Vision \ Wen Wang^*, Canyu Zhao^*, Hao Chen, Zhekai Chen, Kecheng Zheng, Chunhua Shen

project | arXiv | github

Generates text-aligned, identity-consistent storytelling images from stories and character references. Uses language-model planning and dense condition generation to reduce manual control requirements.