BIMBA: Selective-Scan Compression for Long-Range Video Question Answering

Benchmark Model Rank Results
video-question-answering-on-next-qaBIMBA-LLaVA-Qwen2-7B#5Accuracy: 83.73
video-question-answering-on-perception-testBIMBA-LLaVA-Qwen2-7B#2Accuracy (Top-1): 68.51
zero-shot-video-question-answer-on-egoschema-1BIMBA-LLaVA-Qwen2-7B#1Accuracy: 71.14
zero-shot-video-question-answer-on-video-mme-1BIMBA-LLaVA-Qwen2-7B#4Accuracy (%): 64.67
zero-shot-video-question-answer-on-vnbenchBIMBA-LLaVA-Qwen2-7B#1Accuracy: 77.88