The Blending Ratio Is Not Where the Performance Is: Diagnosing Prototype Blending for Few-Shot Adaptation of Vision-Language Models
cs.CV, cs.LG
Submitted: 2026-08-23
Updated: 2026-08-23
Terminology
Sources
- Project and Mix: Task-Semantic Prototypes for Few-Shot Image Classification
- Benchmarking Few-shot Transferability of Pre-trained Models with Improved Evaluation Protocols
- A Survey of Low-shot Vision-Language Model Adaptation via Representer Theorem
- SimpleShot: Revisiting Nearest-Neighbor Classification for Few-Shot Learning
- Cross-Modal Mapping: Mitigating the Modality Gap for Few-Shot Image Classification
- Fine-Grained Visual Classification of Aircraft
- SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features
- DINOv3
Related papers
- Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
- AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
- Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
- MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
- TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
- A Survey on Efficient Vision-Language-Action Models