Semantic Head Specialization Guides Hybrid ViT Attention for Multimodal LLMs
Hybrid attention dominates frontier LLMs, yet Vision Transformers (ViTs) in multimodal LLMs lack a satisfactory hybrid design, with no consensus on why certain attention patterns work better. To fill this gap, we study ViT attention heads and find they differentiate into object- and background-specialist roles, a pattern most pronounced under full attention; we call this Semantic Head Specialization (SHS). We propose SHS-Index to quantify this specialization, show that it distinguishes full-attention from chunk-window ViTs, and find that it strongly tracks downstream benchmark performance. We then identify three structural factors that shape SHS---window interaction, token serialization, and local softmax allocation---and use them as design principles for hybrid attention. Guided by these factors, we design Ariadne Attention, a hybrid that matches full attention on 22 image and video tasks at 6.5x less attention compute. Our findings establish head specialization as a measurable property for diagnosing and designing principled hybrid ViT attention at the multimodal-LLM scale.
Code (0)
등록된 구현이 없습니다.
Similar Papers 제목 키워드 기반
HydraHead: From Head-Level Functional Heterogeneity to Specialized Attention Hybridization
The quadratic complexity of attention poses a critical bottleneck for long-context processing, spurring interest in hybrid attention designs. Most open-source hybrid models adopt a layer-wise strategy. Yet, prior work ha…
Interpreting and Exploiting Functional Specialization in Multi-Head Attention under Multi-task Learning
Transformer-based models, even though achieving super-human performance on several downstream tasks, are often regarded as a black box and used as a whole. It is still unclear what mechanisms they have learned, especiall…
Multi-Task LearningTransfer LearningSpecialization of softmax attention heads: insights from the high-dimensional single-location model
Multi-head attention enables transformer models to represent multiple attention patterns simultaneously. Empirically, head specialization emerges in distinct stages during training, while many heads remain redundant and …
Head Pursuit: Probing Attention Specialization in Multimodal Transformers
Language and vision-language models have shown impressive performance across a wide range of tasks, but their internal mechanisms remain only partly understood. In this work, we study how individual attention heads in te…
Image ClassificationQuestion AnsweringSome Attention is All You Need for Retrieval
We demonstrate complete functional segregation in hybrid SSM-Transformer architectures: retrieval depends exclusively on self-attention layers. Across RecurrentGemma-2B/9B and Jamba-Mini-1.6, attention ablation causes ca…