Reproduce “AdSpark: A Large-Scale Dataset and Benchmark for Product-Centric Advertisement Video Generation” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Zhifei Yang, Zhao Jiang, Keyang Lu, Honghe Zhu, Zheng Zhang, Jingjing Lv, Changping Peng, Ching Law and 1 more

AdSpark-300K and AdSpark-Bench support product-centric advertisement video generation by pairing structured training examples with an ad-specific benchmark. The dataset links product identity, selling points, creative plans, and audio scripts across 300K real and synthetic ads. Fine-tuning LTX-2 raises cross-shot product consistency from 33.30 to 93.59 on a 220-case, SKU-disjoint benchmark, measuring appearance similarity across cuts, not real-world ad performance.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.