VCIFBench: New benchmark for complex instruction following in video understanding

Arxiv CS.CL (NLP/language models)

VCIFBench evaluates multimodal LLMs on video understanding with complex instructions and explicit output constraints. It covers diverse video scenarios to assess models' ability to follow detailed prompts.

multimodalbenchmarks

Arxiv CS.CL (NLP/language models)·Huangchen XuYuan WuYi Chang

7 days ago