Reproduce “ThinkV2V: Unleashing the Reasoning Capability of MLLMs for Instruction-Guided Video Editing” 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
Donghao Zhou, Haoyang He, Fan Zhang, Hao Yang, Guisheng Liu, Xin Gao, Zhongwei Wan, Xingyuan Bu and 5 more
ThinkV2V uses an 8B multimodal language model to reason over a video and instruction, refine the request, and guide a 5B DiT through a connector. It evaluates indirect-target edits on a 308-pair benchmark. Under Seed-1.6-VL, it scores 2.72, versus 2.09 for OpenVE-Edit; the result is benchmark-limited, and does not establish correct reasoning traces or broad real-world generalization.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.