LLaVA-Octopus: Unlocking Instruction-Driven Adaptive Projector Fusion for Video Understanding

arXiv:2501.05067v3 Announce Type: replace Abstract: In this paper, we introduce LLaVA-Octopus, a novel video multimodal large language model. LLaVA-Octopus adaptively weights features from different visual projectors based on user instructions, enabling us to leverage the complementary strengths of each projector. We observe that different visual projectors exhibit distinct characteristics when handling specific tasks. For instance, some projectors excel at capturing static details, while others are more effective at processing temporal information, and some are better suited for tasks requiring temporal coherence. By dynamically adjusting feature weights according to user instructions, LLaVA-Octopus dynamically selects and combines the most suitable features, significantly enhancing the model's performance in multimodal tasks. Experimental results demonstrate that LLaVA-Octopus achieves excellent performance across multiple benchmarks, especially in tasks such as video question answering, long video understanding, and comprehensive multi-choices benchmarks, highlighting its broad application potential.

Related

• Token Reduction via Local and Global Contexts Optimization for Efficient Video Large Language Models
• AdaSpark: Adaptive Sparsity for Efficient Long-Video Understanding
• Reducing Peak Memory Usage for Modern Multimodal Large Language Model Pipelines
• Geometry-Guided 3D Visual Token Pruning for Video-Language Models
• EvoComp: Learning Visual Token Compression for Multimodal Large Language Models via Semantic-Guided Evolutionary Labeling

Source: arXiv cs.CV | 2026-04-21