1 paper
Zhaolu Kang, Shiyu Liu, Tailong Luo +12
Video multimodal large language models (MLLMs) keep climbing video question answering benchmarks, yet shuffling the frames, masking the segment that supports the answer, or occludi…