35 Seeing the Forest and the Trees: Query-Aware Tokenizer for Long-Video Multimodal Language Models Alpachino 129
