Beyond Euclidean Tokens: Hyperbolic Structure-Aware Mapping for Dual-Task Scene Parsing With Only Minimal Trainable Parameters.
The proposed HyperMapper is a hyperbolic structure-aware mapping framework that bridges semantic understanding and geometric priors through hyperbolic token-to-feature interactions, and establishes a promising direction for task-preserving dual-task adaptation, bridging semantic and geometric learning and paving the wa...