Zequn Xie

FarsightAlign: Early-Stage Test-Time Scaling for Prompt-Aligned Text-to-Image Generation

Spot the Key, Recover the Rest: Dual-Path&View Representation Learning for Text-Video Retrieval