Problem
Research questions and friction points this paper is trying to address.
Vision-and-Language Navigation
Efficient VLN
Model Compression
Knowledge Distillation
Query, Align, and Distill: Navigation-Aware Cross-Modal Interaction for Efficient Vision-and-Language Navigation
🤖 AI Summary
This study addresses the parameter redundancy, high computational cost, and deployment challenges of large-scale vision-language navigation (VLN) models by proposing a query bottleneck-based knowledge distillation framework. Methodologically, a teacher model is constructed to explicitly extract navigational evidence, employing learnable query slots to generate global and local navigation representations. Using the query bottleneck as the distillation interface, this work introduces a novel navigation-aware token-adaptive loss function to transfer cross-modal attention and action policies into a compact student model. Experimental results demonstrate that the student model reduces parameters by 93.65% while maintaining navigation performance comparable to the teacher on standard benchmarks, thereby achieving efficient and lightweight deployment.
📝 Abstract
Recent large-scale Vision-and-Language Navigation (VLN) models deliver strong accuracy but remain costly to deploy due to their large parameter counts and computational requirements. We tackle efficient VLN in two steps. First, we build a high-performing teacher that makes navigation evidence selection explicit and compressible. The teacher introduces a small set of learnable query slots to extract global and local action-sufficient navigable evidence from panoramic observations via a Navigable Query Generator, then progressively grounds these evidence tokens to the instruction with an Instruction-Query Aligner for policy prediction. Second, using this explicit query bottleneck as a distillation interface, we train a compact student by transferring both where to attend and what to do. We distill the teacher's global and local navigable queries with a navigation-aware token-adaptive objective, then further match action distributions during fine-tuning. Experiments on standard VLN benchmarks demonstrate that our student nearly matches the teacher's navigation performance while reducing the number of parameters by 93.65% compared to the teacher.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Vision-and-Language Navigation
Knowledge Distillation
Navigable Query Generator
Token-adaptive Objective
Efficient VLN
🔎 Similar Papers
No similar papers found.
💼 Related Jobs
No related jobs found.
