Justin Cui

One-Forcing achieves state-of-the-art one-step video generation while slashing training costs to a third of previous methods.

Scaling visual preference optimization hinges on data quality, as a massive, high-resolution dataset renders complex optimization algorithms unnecessary.

LLMs can guide dense retrieval far more effectively by actively exploring the embedding space with Gaussian Processes, outperforming standard reranking even with a limited LLM budget.