Seeing the Forest and the Trees: Query-Aware Tokenizer for Long-Video Multimodal Language Models Paper • 2511.11910 • Published Nov 14 • 35