CORTEXA
← Browse

Fengbin Guan

1 paper indexed

arxivcs.CV2026-06-29

Goku: A Million-Scale Universal Dataset and Benchmark for Instruction-Based Video Editing

Sen Liang, Cong Wang, Zhentao Yu, Fengbin Guan, Zhengguang Zhou, Teng Hu, et al.

Existing instruction-based video editing datasets commonly focus on single-task appearance editing, failing to meet the complex creative demands of real-world scenarios. To bridge this gap, we present Goku, a large-scale dataset featuring 2 million high-quality, instruction-align…

View free PDFSource page